# AmunCore > AmunCore is a no-code platform that turns any relational database into a secure REST API in minutes, and connects that data to AI. The same endpoints serve apps and partners over REST and GraphQL, AI assistants such as Claude and ChatGPT over MCP (Model Context Protocol), AI agents, dashboards built in App Builder, and your own clients' assistants through Client Connectors, all under one set of keys, permissions, row rules and audit logs. AmunCore is a commercial product of HYNOWorld (Alexandria, Egypt), sold at https://amuncore.com as a managed cloud service or installed on the customer's own servers. It sits in front of an existing database as a controlled access layer: it never copies or exports the customer's records. Key facts: - Databases: SQL Server, MySQL, MariaDB, PostgreSQL, Oracle and SQLite. Endpoints can also wrap stored procedures and external APIs. - REST: GET, POST, PUT and DELETE on /api/v1/{appId}/{endpointName}, with an API key in the X-Api-Key header (or Authorization: Bearer). Paging, sorting and column filters on every list. Writes take an Idempotency-Key and a dry run (?dryRun=true). - MCP server: https://amuncore.com/mcp, signed in with OAuth (no setup) or a connection token in the MCP-Token header. 17 tools, each one switchable per connection. - AI agents: bring your own key for OpenAI, Anthropic, Google Gemini, Groq, Mistral, DeepSeek and others, or use the built-in assistant, which answers in English and Arabic with no AI cost. - Security: tenant isolation, scoped API keys, row and column rules, approval policies, TOTP two-factor sign-in, AES-256 encryption of secrets, full audit log with forwarding to a SIEM. - Plans: Free, Starter, Pro and Enterprise. MCP and AI access need Pro or Enterprise. - Contact: info@amuncore.com This file holds the full text of every AmunCore documentation page listed in https://amuncore.com/llms.txt, so an AI system can read everything in one request. Source: https://amuncore.com/docs/getting-started.md ## Getting started with AmunCore > Turn a database table into a secure REST API in five steps, with no backend code. ### No database yet? Try with sample data On the Dashboard, Try with sample data adds one or more trades (Retail and POS, Distribution, Clinics, Real estate, Restaurants, HR and payroll, Schools, Manufacturing; English or Arabic): their tables as read-only endpoints in one application, a read-only API key, and a ready assistant with its ready questions. The sample data is shared and read only; your own data is never touched. On your own server, the platform administrator builds it once under Demo Database (SQL Server recommended), and its dates move forward by themselves. ### 1. Register an application An application is a connection to one of your databases. In AmunCore open Applications, then New Application, and enter: - Name: a friendly label, for example "Online Shop". - App ID: a short URL-safe id, for example `shop`. It appears in your API URLs. - Database provider: SQL Server, MySQL, MariaDB, PostgreSQL, Oracle or SQLite. - Connection string: how AmunCore reaches the database. It is stored encrypted (AES-256). When you save, AmunCore creates the application's API key. Keep it safe: it authenticates your calls. ### 2. Create an endpoint An endpoint maps a table, a stored procedure, an external API or a Shape to a REST URL. - Name: for example `customers`. - Methods: GET (read), POST (create), PUT (update), DELETE (remove). - Table and columns: load the tables, tick the columns to expose. The primary key is detected for you. - Description: tells AI assistants what the endpoint is for. Recommended. Bulk create makes one read endpoint per table for many tables at once. ### 3. Make the first call ``` curl https://amuncore.com/api/v1/shop/customers \ -H "X-Api-Key: YOUR_API_KEY" ``` The Endpoints page has a Copy cURL button with your real app ID and key filled in, and the API Console tries any endpoint in the browser. ### 4. Filter and page ``` GET /api/v1/shop/customers?Status=active&page=1&pageSize=20 GET /api/v1/shop/customers/42 ``` ### 5. Go further - [REST API reference](https://amuncore.com/docs/rest-api.md): methods, filters, errors, GraphQL-style queries and webhooks. - [AI access over MCP](https://amuncore.com/docs/mcp.md): let Claude, ChatGPT or another assistant use the same endpoints. - [App Builder](https://amuncore.com/docs/app-builder.md): dashboards and report pages from your endpoints. - [AI agents](https://amuncore.com/docs/agents.md): scheduled reports, alerts and chat apps. - [Security and access control](https://amuncore.com/docs/security.md): keys, row rules, approvals and audit. Source: https://amuncore.com/docs/endpoints.md ## Building endpoints in AmunCore > An endpoint is a REST URL in front of a table, a stored procedure, an external API or a Shape. Every endpoint gets the same API keys, permissions, row rules, rate limits and audit trail. ### Table endpoints (visual builder) - Load the database's tables and pick the columns to expose. The primary key is detected for you. - Add joins, filters and combined columns, with a live SQL preview. - Test the connection before saving. - Clone an endpoint to make a similar one quickly. - Bulk create: tick many tables and get one read endpoint per table in one step. - ERP templates: ready-made mappings for Microsoft Dynamics, SAP, Oracle, QuickBooks, Xero, Zoho and more. - Every query is parameterized, so values never become SQL. ### Stored procedures Expose an existing stored procedure as an endpoint. Input parameters come from the query string (GET) or the JSON body (POST); output parameters and result sets are returned as JSON. ### External APIs (passthrough) An endpoint can forward the request to an API you already use, such as a partner service, a payment provider or another department's system. Callers use one AmunCore key instead of the upstream credentials, and the call gets AmunCore's limits and audit log. Destinations must first be approved on an allowed-hosts list. APIs that answer in XML are supported. ### Shapes: fixed document formats A Shape produces the exact document a partner or authority requires (nested objects, totals, formulas, checks) from tables you cannot change. - Build it in a designer, with filters, sorting and values passed by the caller (for example `?id=INV-1`). - Check the result against the required structure before publishing. - Answer in JSON or XML, for example UBL e-invoices in the UAE. - Frozen copies keep each document byte for byte with its SHA-256 fingerprint, the Shape version, the values passed and who asked, as proof of exactly what was sent. - The formula assistant is built in and needs no outside AI service. ### Endpoint descriptions A clear description on each endpoint tells AI assistants what it holds, so they pick the right one. It also appears in the generated API documentation. ### Live feed for Excel and Power BI An endpoint that reads (GET) can have a live feed: a secret link that Excel and Power BI read and refresh, with no key to type. In Endpoints, press Feed, then Make a link. Two links are shown once: CSV for Excel (Data, From Web) and JSON for Power BI (Get data, Web, Anonymous). The feed reads with the access of the person who made it (row and column rules), up to 20,000 records, at most 240 reads an hour; it stops when that person is switched off. A new link replaces the old one at once; Switch off deletes it. Only a hash of the link is kept. Deleting an application (its Edit page, Delete Application) first lists, by name, everything that goes with it: its endpoints, the assistants and agents that use it (also only in their steps), the Client Connectors that share it, the App Builder apps that show its data, documents or data files (or whose Ask block talks to an assistant that goes), its live feed links, and anything an earlier delete left behind (marked left over). Type its name to confirm; a loader shows each step while it is deleted. Your database itself is never touched. Restoring: a deleted application can be restored for 7 days (Applications, Recently deleted). Every name and address is checked first (application name and API address, assistant, App Builder app and Client Connector names and web addresses); anything used now gets a free one, shown before restoring. Live feed links, open client sign-ins, run history and usage counts are not restored. Delete for good removes a kept application at once. ### Related - [REST API reference](https://amuncore.com/docs/rest-api.md) - [Security and access control](https://amuncore.com/docs/security.md) - [AmunCore overview](https://amuncore.com/llms.txt) Source: https://amuncore.com/docs/rest-api.md ## AmunCore REST API reference > Every AmunCore endpoint is a REST URL under /api/v1/{appId}/{endpointName}, protected by an API key, with paging, sorting, column filters and a consistent JSON response and error format. ### Authentication Send the application's API key with every request, in either header: ``` X-Api-Key: YOUR_API_KEY Authorization: Bearer YOUR_API_KEY ``` Scoped keys can be read-only, write-only or full access, with an expiry date. See [Security and access control](https://amuncore.com/docs/security.md). ### Operations | Method | URL | What it does | |---|---|---| | GET | /api/v1/{appId}/{endpointName} | List records | | GET | /api/v1/{appId}/{endpointName}/{id} | Get one record | | POST | /api/v1/{appId}/{endpointName} | Create a record | | PUT | /api/v1/{appId}/{endpointName}/{id} | Update a record | | DELETE | /api/v1/{appId}/{endpointName}/{id} | Delete a record | Each method works only when the endpoint allows it. ### Filtering, paging and sorting ``` GET /api/v1/shop/customers?Status=active&City=Cairo GET /api/v1/shop/customers?page=2&pageSize=50&orderBy=Name&orderDir=asc ``` - Column filters: `?ColumnName=value`. - Paging: `page` and `pageSize`. - Sorting: `orderBy` and `orderDir` (`asc` or `desc`). Paging starts when `page` is sent: `pageSize` is 20 by default and at most 500. Without `page`, every record you may see is returned. ### Responses Reads: ```json { "success": true, "data": [ { "Id": 1, "Name": "Acme Co" } ], "meta": { "endpoint": "customers", "method": "GET", "durationMs": 9, "timestamp": "2026-10-27T10:30:00Z" } } ``` Writes return the record as the database wrote it (POST answers 201): ```json { "success": true, "data": { "Id": 43, "Name": "Nile Foods" }, "meta": { "endpoint": "customers", "method": "POST", "durationMs": 18, "timestamp": "2026-10-27T10:30:00Z" } } ``` ### Writing safely - **Idempotency-Key**: send `Idempotency-Key: ` with POST, PUT or DELETE. A retry with the same key and body returns the first answer (`"idempotentReplay": true`, header `Idempotency-Replayed: true`) and writes nothing. The same key with a different body answers 409. Keys are kept 24 hours. - **Dry run**: add `?dryRun=true` (or `X-Dry-Run: true`) to check a write without doing it. Every check runs; a change or delete says how many records it would touch; nothing is written, held for approval or sent to webhooks. Not available for stored procedures or external APIs (400). - **Approval**: an endpoint with an approval policy answers 202 with `approvalRequestId`; nothing is saved until an approver accepts it. - **Undoing**: a created record can be removed with DELETE; a change can be undone with a PUT of the previous values (read them first). A DELETE cannot be undone: AmunCore keeps no copy of your records. ```json { "success": true, "data": { "dryRun": true, "action": "update", "fields": ["City"], "id": "42", "matches": 1, "message": "Checked: 1 record would be updated. Nothing was saved.", "approvalNeeded": false } } ``` ### Errors Every error has the same JSON shape: `success` (false), `message` (readable), and where relevant `error` (a short code), `requestId` and `timestamp`. | Status | Meaning | |---|---| | 400 | Bad request | | 401 | Missing or invalid API key | | 403 | The key may not do this (read-only key, IP allow list) | | 404 | Not found | | 405 | The endpoint answers another method | | 409 | Idempotency-Key used with a different body, or the same request still running | | 422 | An endpoint that answers with a Shape: the document did not pass its checks | | 429 | Too many calls (the plan's allowance, or the endpoint's calls per minute); see Retry-After | | 500 | Server error | A change sent to an endpoint with an approval policy is held for approval instead of being saved straight away. ### GraphQL-style queries Fetch several endpoints and choose exact fields in one request: ``` POST /graphql/{appId} X-Api-Key: YOUR_API_KEY { "queries": [ { "endpoint": "customers", "fields": ["Id","Name","Email"], "limit": 10 }, { "endpoint": "orders", "fields": ["Id","Total","Status"], "limit": 5 } ] } ``` Only fields the endpoint exposes can be asked for. ### Webhooks AmunCore can POST to your URL when records are created, updated or deleted, with timeouts, retries and custom headers. Each delivery carries: - `X-AmunCore-Signature`: HMAC-SHA256 of the body, to verify it came from AmunCore. - `X-AmunCore-Delivery`: a unique delivery id, to ignore duplicate retries. - `X-AmunCore-Attempt`: the attempt number. ### Limits Every answer of a plan with a call allowance carries `X-RateLimit-Limit` (the plan's calls) and `X-RateLimit-Remaining` (calls left). An endpoint can also allow only so many calls a minute. Either limit answers 429 with `Retry-After`. ### Versions and status The API is versioned in its path (`/api/v1`). Every release is in the [changelog](https://amuncore.com/docs/changelog.md). [Service status](https://amuncore.com/status). ### OpenAPI - [OpenAPI 3.0 specification](https://amuncore.com/openapi.json) (public), with examples and code samples for every operation - [Swagger UI](https://amuncore.com/swagger) Each application also gets generated, shareable documentation for its own endpoints, with code samples in cURL, JavaScript, Python, C# and PHP. Source: https://amuncore.com/docs/mcp.md ## AI access over MCP > AmunCore is an MCP (Model Context Protocol) server. Claude, ChatGPT and any MCP client can list your applications and endpoints, read and write through them, work out totals, find records and read your documents, with the same permissions, row rules and audit log as the REST API. ### Connect an assistant The MCP address is `https://amuncore.com/mcp` (on your own servers, your own address followed by `/mcp`). An admin first switches on AI access on the AI access page. 1. OAuth connector (recommended, no setup): in Claude (desktop, web or mobile) open Settings, Connectors, Add custom connector, enter the MCP address and press Connect. Sign in to AmunCore and press Allow. The assistant then works as that person, with their own access. 2. Connection token: for ChatGPT, scripts or any HTTP client, create a connection on the AI access page and send its token in the `MCP-Token` header with each request. 3. Older Claude Desktop versions can use the mcp-remote bridge (needs Node.js); the AI access page shows the exact configuration. MCP clients discover sign-in through [the OAuth protected-resource metadata](https://amuncore.com/.well-known/oauth-protected-resource). ### Tools Assistants discover tools with `tools/list` and call them with `tools/call`. Each tool declares its parameters and returns structured JSON. | Tool | What it does | |---|---| | list_applications | List the company's applications | | get_application | Details of one application | | list_endpoints | List the endpoints of an application, with their descriptions | | call_api | Read or change data through any allowed endpoint (GET, POST, PUT, DELETE) | | get_totals | How many, total, average, lowest, highest or how many different values, by a field or by month, day or year, with conditions, worked out by the database over every record (read only) | | find_records | Records with conditions and in an order, for example the 10 largest unpaid invoices, chosen among every record, with how many match (read only) | | list_files | List stored documents | | get_file_text | Read a document's text | | search_files | Search documents by name and content | | get_dashboard_stats | Overview figures | | create_application | Register a new application or database | | create_endpoint | Add an endpoint to an application | | toggle_endpoint | Switch an endpoint on or off | | regenerate_api_key | Make a new API key (the old one stops working) | | get_audit_logs | Read recent API and AI activity | | get_license_status | Check the plan and its limits | | check_plan_limits | Check whether more can be created | The company chooses which tools its assistants may use. For everyday use the read tools plus call_api are recommended; get_totals and find_records are off until ticked. ### What each connection can be limited to - Applications and endpoints: all, several or one. - Tools: exactly which tools are offered. - Read-only: write requests are refused. - Who can use it: everyone in the company or chosen people. - Expiry date, calls per minute and allowed IP addresses. - Active or disabled, without deleting it. ### Safety - Every tool call goes through the endpoint as the caller: permissions, row rules, column rules, approval policies and fixed filters apply exactly as for a REST call. - Writes can require named fields first and a human approval before anything is saved. - Text from records and documents is treated as data, never as instructions. - Every call is recorded in the audit log, and AI access can be switched off for the whole company at once. ### Related - [Client Connectors](https://amuncore.com/docs/client-connectors.md): the same for your own clients' assistants. - [AI agents](https://amuncore.com/docs/agents.md): AmunCore's own agents and assistant. - [AmunCore overview](https://amuncore.com/llms.txt) Source: https://amuncore.com/docs/agents.md ## AI agents in AmunCore > An agent is a worker that uses your endpoints and documents: it answers questions, runs scheduled reports and alerts, and can be published as a chat or form app. It thinks with an online AI using your own key, or with AmunCore's built-in engine at no AI cost. ### Two ways an agent can think - Online AI with your own key: OpenAI, Anthropic (Claude), Google Gemini, Groq, Cerebras, DeepSeek, Mistral, xAI (Grok), Z.ai (GLM), Moonshot (Kimi), Alibaba (Qwen), OpenRouter, Azure OpenAI, or any service that speaks the OpenAI chat format (including Ollama or LM Studio on your own servers). The provider bills you directly. - Built-in engine (no AI): answers common questions from your records and documents in English and Arabic (including Gulf, Levantine and Egyptian phrasing), and runs steps you choose: get records, group and total, stop if, email, save, show a message. Free and instant. ### What the built-in engine understands - Lookups: "order 1001", a name, a code. - Filters: "orders with status cancelled", "unpaid orders this month". - Counts, totals and averages, by group and by period, compared with the same days of the period before (each group too). - Largest, smallest, newest and oldest records, chosen by the database among every record. - A daily briefing of each list, and questions in two parts ("sales today and unpaid invoices"). ### Templates Grouped by what they do: Answer questions (Ask your data (no AI), Ask your data with AI, Document Q and A, Policy agent (no AI), Search chat (no AI), Public product agent), Reports and alerts (Daily report by email, Weekly summary with AI, Low stock alert, Overdue payments, Data quality check) and Forms and data entry (Request form, Lookup form, Data entry agent). Filter them by No AI or Online AI. ### Ready questions Questions people can ask the agent, ready to press. "Make questions from the data" reads a sample of what the agent can use (its ticked applications and endpoints, its documents' Excel and CSV files, or its Get records steps), works on the page as it is (also before saving), and writes questions that fit each list: how many, the total this month, this month against last month, totals by a group, customer or month, the top 5, anything unusual, the unpaid ones, the latest, an analysis, and the ageing report, collection list and cash forecast when the data has invoices and dates. Ready questions can be written in Arabic. The questions people ask most come first. An industry pack (28 trades, from retail and clinics to manufacturing, schools, hotels, logistics, lending, insurance and law firms) adds that trade's questions where the data has their lists and columns, and its words as business words pointing at the agent's own lists. People add, change or remove them (up to 40). "Test them" (built-in engine) asks each as a test, nothing kept, and shows a score such as "Answers 18 of 19". They replace the check questions of earlier versions. They are also the team's questions: in the app's side panel, by group, next to each person's own Saved and Recent, to ask at any point of a chat. A company admin using the app adds one under any answer (Add to team questions), into a group they pick or a new one. ### Schedules and emails Run every few hours, day, week or month at a time in the agent's own time zone. Results can be emailed as a formatted report, with a bar chart under grouped tables. The built-in engine writes an analysis of a list when asked ("I need details with analysis for invoices", "حلل الفواتير"): what stands out, the figures, the trend by month, the top names and their share, what is owed and late, the largest records, and how it worked it out. The built-in engine answers the questions a finance manager asks, from the lists it can read: - "What are late payers costing me", "cost of the top 10 customers paying more than 30 days late", "تكلفة التأخير": what is owed past the due date, by customer, and what carrying it costs (unpaid amount x yearly rate x days late / 365; say a rate, "at 25%", or it uses 20% and says so). - "What should we sell more of", "which clients am I losing money on", "ربحية المنتجات": revenue less direct cost by product or by customer, with what to do for each: Grow, Review, Reprice, Investigate or Data missing. A customer's cost of paying late is taken off too. - "What is shrinking working capital", "why is there never cash", "رأس المال العامل": days to collect (DSO), days in stock (DIO), days to pay (DPO), the cash cycle, and the customers and items holding the most cash. - "Ageing report", "aged receivables", "أعمار الديون": what each customer owes by days past the due date (not due yet, 1-30, 31-60, 61-90, over 90), with the totals of each. - "Collection list", "who should I call first", "قائمة التحصيل": the customers to chase, the costliest first (amount overdue x days late), with what they owe, their oldest invoices, their phone and email (from the customers list) and what to do. - "Cash forecast", "cash flow for the next 13 weeks", "توقع السيولة": cash expected in from customers (each invoice on its due date moved by that customer's usual delay; overdue up to 90 days spread over the next 4 weeks; older left out as at risk) and out to supplier bills, week by week; add today's cash ("starting with 250000") for the balance each week. The cost of a product can sit on the invoice lines or as a cost price in the products list; the customer can sit on the line or on its invoice. When a list or column is missing, the answer says "Data missing" and names it. A list longer than the agent reads is worked out by the database over every record. The agent's Yearly cost of money (under Its job) sets the rate for late payment; a rate in the question wins. The lists can also be data files: tick Documents for the application that holds them, and the built-in engine reads its Excel, CSV, JSON and XML files as records, with no step to add (the files the question names, or for a finance question those named like invoices, customers or bills; each sheet of a workbook is its own list). So a company with only an invoices sheet can ask "what are late payers costing me". Invoices that keep only a CustomerID are answered by customer name from the customers list; amounts in several currencies are answered one currency at a time ("what are late payers costing me in AED"). The built-in engine learns questions: when it cannot answer one, teach what it means (built from menus in the test console, or written, checked first), and it is answered that way from then on, also with other customers, cities, numbers or periods ("money stuck at Delta Trading last month" = "what does Delta Trading owe last month" also answers "money stuck at Nile Foods this week"). When it cannot answer, it offers up to 3 questions under "Did you mean"; with "Learn from what people choose" ticked, the one pressed is remembered. What people said lists the questions it could not answer and those marked wrong, ready to teach, and the taught ones; the weekly summary lists the questions not answered. The built-in engine can answer a question on each run, such as "brief me" for a morning briefing, "unpaid invoices this month" or "sales by branch this month", and email the answer. ### Publish as an app - Alerts in chat (v3.55.1): people set alerts in words in their own Telegram or WhatsApp chat ("tell me when today's sales are below 50,000"); the agent replies with the figure now and asks "Set this alert?" (Yes or No); checked as that person, sent only when the condition becomes true; "my alerts", "stop alert 2". Off until an admin switches them on (Channels page), where they also choose who can set them (team and invited people, or anyone who can chat with it). Works with offline and online AI agents. v3.55.2: also new records ("tell me when there is a new booking for Mohamed Samy", "قولي لما يكون في حجز باسم محمد سامي": it writes when there is a new one and shows it), a change, a rise or a fall ("tell me when the price of rice changes"), and "just once". v3.57: reminders ("remind me in 3 days to call Nile Foods", "ذكرني بكره الساعة 9 ابعت التقرير": a Yes first, then one message at that time, up to 90 days ahead), and alerts for someone else: a person who connected their chat ("tell Ahmed when there is a new booking") or, for new records, the person a column names ("when a new booking arrives, message the doctor", "لما يوصل حجز جديد ابعت للدكتور"). Only the team can set alerts for others. v3.57.1: chat invitations (Channels page, Invite tab: each person gets an email with their own one-time links and QR codes, for 48 hours) and QR codes in the app and for apps for anyone. - Agents dashboard (v3.53.3): questions by channel, the latest questions (every 30 seconds), where each main agent sent its questions, those not answered (with Teach), answer times, who asks most and busy hours; filters by agent and channel. Long lists show in pages of 10 (v3.54). - In every chat app (v3.53.2): a total as a number card; a trend by month or a top list shows its chart first; Ask next under an answer (the same for the period before, and others of the same list); dark mode (kept in that browser); a language button (English or Arabic) for an app set to Automatic. - A main agent's app: a card for each agent it covers with its questions; Ask above the box chooses who answers (Any: it chooses); each answer under the agent's name, via the main agent. - Questions and chat (the ready questions as buttons grouped by list, narrowed by a period and up to 3 lists of values such as Customer or City, the answer as a report, and a box to type any question) or Form (boxes and one button: look up a record, or send a request saved through your API). - The built-in engine needs no steps to answer questions: it reads the lists it can use that a question is about. Steps are for a job (a report by email, a saved record) and use only what it can use. - A new agent is made step by step, each part checked, with a review before Create; or all on one page. - For your team: people sign in and the agent works as each person, with their own rules. - For people you choose (their emails) or the invited clients of a Client Connector: they sign in with a 6-digit email code; read only; clients read as themselves. - The coverage score on each agent's Activity page: the share of people's questions it answered, week by week. - For anyone: no sign-in, always read only, with a limit on runs per hour from each address. - App language: automatic, English or Arabic. - Chat button on your website: one line (``) adds a chat button on the websites you list (https, up to 5). An app for anyone opens in a panel on the page; an app that signs in opens in a new tab. No other website can show it. - Short answer (built-in engine): the answer in words with figure cards; each list as a card with Excel, Word and PDF files made in the browser. People can switch to the full report. ### Main agent Main agent (v3.53, its own kind since v3.53.1): choose Main agent under Basics; its steps are Basics, Brain, Agents (the person it works as and the agents to tick), App, Advanced, Review, with no data, ready questions, steps or schedule of its own. Each question goes to the agent it is about (by its name, description, ready questions, business words, instructions and data names, English and Arabic; when an Arabic and an English agent match the same, the question's language chooses; a follow-up stays with the last one; unclear: an AI main agent asks its model, else buttons to choose). A question about two agents is answered by both. Its app decides which it can use: team agents only when its app is for your team; public ones always; ones for chosen people or clients never (they answer only in their own app); the page marks the ones it cannot use and saving is refused until fixed. A team person reaches team agents (as themselves) and public ones (read only); others only public ones. It cannot send to itself or another main agent. Answers are labelled with the agent's name and kept on its history. One app, Telegram bot and WhatsApp number for all. ### Telegram and WhatsApp A chat app's agent can also answer on Telegram and WhatsApp (agent page, Channels): the same answers, people and rules as its published app. Telegram is free: paste a bot token from @BotFather. WhatsApp uses Meta's WhatsApp Business Platform: paste the phone number ID, a permanent access token and the app secret, then paste the callback URL and verify token AmunCore shows into Meta's webhook settings. An app for anyone answers anyone, read only; for other apps each person connects their chat once from the app (Use on Telegram or Use on WhatsApp, a one-time code) and is checked again at every message. Only typed messages in a person's own chat are answered. Long lists also come as a CSV file. Secrets are kept encrypted; every message is checked to come from Telegram or Meta. Reports in an app for anyone (v3.51.1): the Channels page lists under Chats the people who wrote to the bot privately, each with a Reports switch (they must have written first). Save on every step (v3.51.1): the agent page has Save on every step; if anything needed is missing, nothing is saved and that part opens at the box to fix. Send now (v3.51): in the Schedule, runs once and sends the result by every way ticked, then says where it went or why not; each send is kept in the run's steps. ### Changes from chat v3.56: an agent with Allow changes saves only through an endpoint that writes (POST adds, PUT changes, DELETE removes); with only lists that read it cannot change anything. - Always a Yes first: in the app, the website chat, Telegram and WhatsApp, AmunCore shows exactly what will be saved with Yes and No; only Yes saves it, within 10 minutes; any other message drops it; Yes twice saves once. - Then the endpoint's rules as the person who asked (required fields, row access, approvals, plan, audit log). The reply says Saved, Sent for approval (request 12), or why not. - Approvals in chat (Channels page, Approvals tab, off by default): approvers who connected their own chat with the agent get each waiting change with Approve 12 and Reject 12 (a reason is needed; Arabic: موافقة 12, رفض 12 السبب). The person who asked in a chat hears the decision there. WhatsApp: within 24 hours of the approver's last message. - The built-in engine (no AI, v3.56.1): "add a booking", "احجز موعد": it asks for each of the endpoint's Required fields, then shows the record with Yes and No. "field: value" in the message is taken as it is. "change booking 88: status: Confirmed", "delete booking 88". "cancel" (إلغاء) stops. - Undo (v3.56.1): after an add, Undo (تراجع) removes the record within 5 minutes, through a list of the same table that deletes. - Cancel a request (v3.56.1): "Cancel request 12" (إلغاء الطلب 12) in the chat it was asked in, while it waits. - Scheduled runs and forms save as before. ### Safety - Everything goes through the REST API as the agent's person: plan quota, row and column rules, approvals and audit log apply. - Changes wait for approval by default. - An agent uses only the endpoints you ticked and emails only the addresses you listed. - Limits per agent: tool calls per request, seconds and tokens per run, runs per day, and an end date. ### Plans Agents need a plan with MCP (Pro or Enterprise). Without numbers in the license, Pro allows 5 agents and 3,000 runs a month, Enterprise 100 agents and 100,000 runs a month. ### Related - [App Builder](https://amuncore.com/docs/app-builder.md) - [AI access over MCP](https://amuncore.com/docs/mcp.md) - [AmunCore overview](https://amuncore.com/llms.txt) Source: https://amuncore.com/docs/app-builder.md ## App Builder (AmunCore Studio) > App Builder makes dashboards, report pages and data pages from your endpoints and data files, with no code, and publishes them for your team, people you choose, your clients or anyone. Every viewer sees only what their own access allows. ### Three kinds of app | Kind | What it shows | |---|---| | Dashboard | Figures (totals, counts, averages) and charts at a glance | | Report page | Charts and a table with filters, to read and download | | Data page | Searchable tables with filters, to find records | The kind only sets the first layout; any block can go on any app. ### Ways to make one - From your data (no AI): choose endpoints and AmunCore lays out figures, a chart by month, charts by group, a table and filters. - With AI (every plan): describe it in your own words, using the provider and key of one of your Online AI agents, or a key pasted for that one request (used once, not saved). - From 57 templates in 15 kinds of business: sales, finance, accounting, stock, shops and restaurants, HR, operations, support, logistics, schools, clinics, property and hotels, marketing, factories, and client or public apps. - With the sample data: on the Dashboard, Try with sample data makes an app for each trade added (Sales, Products, Collections, Customers... and an Ask page with its agent), published for your team, in Arabic and right to left for the Arabic sets. - Column names (Theme, v3.41): the name shown for each column in table headings, figure labels, titles made from columns, exports and email reports; titles you wrote stay as written. - Words on buttons (Theme, v3.40): Arabic puts the published app's own words in Arabic (Refresh, More, filters, notes under tables, chart switches, Ask), and its email reports and alerts too. Titles, column names and data stay as written. ### The designer - Blocks: Figure, Chart, Table, Pivot, Text, Picture and Agent, dragged onto the page. - Charts: bars, columns, line, area, donut, pie, stacked bars split by a second column, and bars with a line for two values; a figure's target as a bar or a gauge. - Arrange with the mouse or keys: beside, above or below another block; choose several with a box and put them in one row, spread them evenly or give them the same size. - Data from endpoints or from data files (Excel, CSV, JSON, XML). - Filters, search, and press-to-filter on charts. - Theme with colors, fonts and motion (Off, Subtle or Lively): blocks come in one after another, charts grow and figures count up, never for viewers who asked their device for less motion. ### Large data Up to 5,000 records are read per endpoint on a page. Above that, figures, charts and pivots are worked out by the database over every record, with the app's filters and the viewer's own access. ### Sharing and use - Who can open it: your team (each person's own access), up to 200 people chosen by email (sign-in with an emailed code, no account), the clients of a Client Connector, or anyone. - Links that keep the page, the filters and what was pressed. - TV mode for office screens, turning pages by itself and refreshing figures. - Download as PDF (blocks keep their widths and rows on paper; landscape or portrait), Excel or CSV. - Email reports every day, week or month with figures, charts and the first rows of tables, to people of your company, invited clients, an app's email list or email addresses you type. - Publish keeps the last 30 versions; unpublish closes the app at once. ### Right to left, titles and conditions - **Right to left (Arabic)** under Theme turns the whole app, its tables, charts and email reports; each block can choose its own direction. - A block's title can show a column's value: *Balance of {AccountName}* follows the filters. - Conditions and color rules take several values separated by commas: *Code is 555, 777, 88*. ### Finance Three templates answer a finance manager's questions. Each has figures and charts from your data and a report the built-in engine writes when the page opens: - **Late payers and what they cost**: what is still owed and what is more than 30 days past due, by customer and month, oldest first; the report says what carrying it costs at a yearly rate. - **Profit by product and customer**: sales, direct cost, contribution and margin by product and customer, with what to do with each (Grow, Review, Reprice, Investigate, Data missing). - **Working capital and cash**: what customers owe, stock, what you owe suppliers; the report gives the days to collect, in stock and to pay, and the cash cycle. - **Ageing and collections**: what is owed, not due yet, past due and over 90 days; the reports give the ageing by customer and the collection list (who to call first, their oldest invoices, how to reach them, what to do). - **Cash forecast**: what is owed to you and to suppliers; the report gives the cash in and out week by week, and the lowest balance when asked with today's cash. An Agent block can answer its first question when the page opens (built-in engine only), and is then printed with the page. In conditions, a date can be said from today: today, today-30, today+7. ### Accounting Four templates work from the journal lines of an ERP or accounting system (date, account code and name, debit and credit; account type, journal number, description and branch if you have them): - **Trial balance**: each account's opening balance, debits, credits and closing balance for the dates chosen, with totals and the difference between debits and credits. - **Account statement (ledger)**: an account's entries with a running balance from its opening balance to its closing balance. - **Income statement and balance sheet**: revenue, expenses and net profit; assets, liabilities and equity at the end of the dates. - **Journal analysis**: debits and credits by month, the accounts used most, activity by branch, and journals that do not balance. Any app can use the same parts: a pivot with several values side by side (each over the dates chosen, before them or up to their end), figures that are opening or closing balances, charts of a formula, and tables with a running balance. With more than 5,000 lines, the database works them out over every line. ### Alerts Watch a figure and email people when it passes a line you set, for example "Unpaid invoices above 100,000" or "Stock below 50". Checked every few hours or once a day, for up to 50 people of your company, each with their own access, for everyone on the email list of an app for people chosen by email, or for email addresses you type (they get the email only, without access to the app). One email when the line is passed; another only after the figure comes back and passes it again. ### Notes Switch on notes and people can write a short note on a page or a figure ("Sales dropped because the Dubai branch was closed"). Everyone who can open the app sees them, with who wrote them and when. Notes are offered in apps for your team and for people chosen by email. ### Related - [AI agents](https://amuncore.com/docs/agents.md) - [Client Connectors](https://amuncore.com/docs/client-connectors.md) - [AmunCore overview](https://amuncore.com/llms.txt) Source: https://amuncore.com/docs/client-connectors.md ## Client Connectors > A Client Connector lets your own clients use their AI assistant (such as Claude) on the data you share with them: their orders, invoices, statements or balances. Clients need no AmunCore account, and each sees only their own records. ### Two kinds | Who can use it | How they get in | Good for | |---|---|---| | Invited clients | Their email and a 6-digit code sent to it; no password | Anything that belongs to one client: orders, invoices, statements, balances | | Anyone | No sign-in, read only | Product catalogue, prices, stock, branches, opening hours | ### Set it up 1. On the AI access page press New connection and choose Your clients or Anyone. 2. Give it a name and an address, for example `https://amuncore.com/c/acme-clients/mcp`. 3. Tick what it shares: whole applications or chosen endpoints. 4. For invited clients, add each client with their own values, for example their customer number. A whole email domain can be allowed too (public email services are refused). 5. Optionally let the AI work out totals and find records (get_totals and find_records), off by default. ### What the client does In their assistant they add a custom connector with your connector's address, press Connect, type their email and the code from the email, then ask in their own words, for example "show my open orders". ### Safety - Clients never get an AmunCore login; their sign-in works only on their connector. - Each client's values are applied to every call, so they see only their own records. - Every call counts toward the plan's monthly calls, a monthly limit can be set per connector, and every call is audited. ### Plans Client Connectors need a plan with MCP. Without numbers in the license, Pro allows 3 connectors with up to 1,000 clients each, Enterprise 50 with up to 10,000 clients each. ### Related - [AI access over MCP](https://amuncore.com/docs/mcp.md) - [App Builder](https://amuncore.com/docs/app-builder.md): apps for the clients of a connector. - [AmunCore overview](https://amuncore.com/llms.txt) Source: https://amuncore.com/docs/files.md ## Files and documents > AmunCore stores your documents per application and lets AI assistants search and read them, while you decide which people, and which AI, can see each file. ### Supported files | Type | Text read for AI | |---|---| | PDF | Yes; scanned PDFs are read with OCR (English and Arabic) where OCR is installed | | Word (.docx) | Yes | | Excel (.xlsx) | Yes, cell values across sheets | | Text (.txt, .csv, .md and similar) | Yes | | Other files | Stored and downloadable only | Executable and script files are blocked. Files can be uploaded one by one or a whole folder at once, with a category such as Contracts or Invoices. ### Who can see a file - Everyone in my company (the default): every member, and the AI, can use it. - Only specific people: visible only to the people chosen and the owner, and hidden from the AI for anyone else. - AI visibility can also be switched off for each file. ### Using files with AI Over MCP, an assistant can list files, search them by keyword and category, and read a file's text, for example "search my contracts for the termination clause". Each person's assistant sees only the files that person may see. Text inside a document is treated as data, never as instructions. Data files (Excel, CSV, JSON, XML) can also feed App Builder apps like an endpoint. ### Plans Files can be stored on every plan. Reading files with AI needs Pro or Enterprise. See [Plans and limits](https://amuncore.com/docs/plans.md). Source: https://amuncore.com/docs/security.md ## Security and access control > AmunCore is a controlled access layer in front of your data. Every channel (REST, GraphQL, MCP, agents, apps and webhooks) goes through the same keys, permissions, row and column rules, approvals and audit log, and each company is isolated from every other. ### Hosting - Self-hosted: installed on your own servers, behind your firewall. Data never leaves your environment. - Managed cloud: AmunCore's hosted service with strict per-company isolation. In both, AmunCore never copies, exports or stores your database records. ### API keys - A primary key per application, plus scoped keys: read-only, write-only or full access. - Expiry dates, revoke at any time, rotate safely. - Optional IP allow list and CORS settings (Pro and Enterprise), and rate limits. ### Who sees which rows and columns - Row rules: limit an endpoint by up to 5 columns (for example branch or company). Each person, API key and AI connection sees only the rows for its own values. - Column rules: hide chosen columns from chosen callers. ### Approvals and AI writes - Approval policies hold a change until a responsible person approves it, for every caller including AI. Approved changes are saved once, in the name of the person who asked. - Required fields stop an assistant saving a record with details missing. ### Accounts - Roles: Viewer, Developer and Admin. - TOTP two-factor sign-in with recovery codes, sign-in lockout after repeated wrong passwords, and a list of signed-in devices that can be ended. - Saved secrets (connection strings, Telegram and WhatsApp tokens, AI keys) are encrypted. The encryption keys are kept in AmunCore's database and can be locked with a secret kept outside it (Security:KeyEncryptionSecret, at least 32 characters, or the variable AMUNCORE_KEY_SECRET), so a copy of the database alone opens nothing. - Assistant app pages send a strict Content-Security-Policy (scripts only from the site and the page's own). - Google and Microsoft sign-in. ### Encryption and queries - Passwords hashed with BCrypt; connection strings and secrets encrypted with AES-256. - Every database query is parameterized, which protects against SQL injection. - Webhooks are signed with HMAC-SHA256. ### Audit - Every API and AI action is logged, filterable and exportable. - Events can be forwarded as they happen to Splunk, Microsoft Sentinel, Datadog, Elastic or any HTTPS event collector. ### Related - [Security and data page](https://amuncore.com/security.html) - [AI access over MCP](https://amuncore.com/docs/mcp.md) - [AmunCore overview](https://amuncore.com/llms.txt) Source: https://amuncore.com/docs/plans.md ## Plans and limits > AmunCore has four plans: Free, Starter, Pro and Enterprise. AI features (MCP, agents, Client Connectors, reading files with AI) need Pro or Enterprise. | Feature | Free | Starter | Pro | Enterprise | |---|---|---|---|---| | Applications | 1 | 3 | Unlimited | Unlimited | | Endpoints per application | 2 | 20 | Unlimited | Unlimited | | API calls | 1,000 a month | 100,000 a month | 5 million a month | Unlimited | | File storage | 50 MB | 1 GB | 10 GB | Unlimited | | Databases | SQL Server | SQL Server and MySQL | All supported | All supported | | CORS and IP allow list | No | No | Yes | Yes | | MCP and AI | No | No | Yes | Yes | | Shapes | Yes | Yes | Yes | Yes | | App Builder apps | 1 | 5 | 50 | Unlimited | | App Builder: who can open apps | Your team | Your team, people chosen by email | Also clients and anyone | Also clients and anyone | | App Builder: email reports | No | Yes | Yes | Yes | | App Builder: alerts | No | No | Yes | Yes | | App Builder: TV mode | No | Yes | Yes | Yes | | Multi-tenant SaaS use | No | No | No | Yes | On your own servers the Free plan allows 1,000 calls a day. App Builder templates, Create with AI (with an Online AI agent or a key pasted for one request) and PDF and Excel downloads are on every plan. A license key can set its own App Builder numbers. When a plan changes, apps are kept: past the number no new one is made, an app shared in a way the plan no longer includes opens only for the team, and reports, alerts and TV mode wait. ### License keys A paid plan is unlocked with a signed license key, activated by an Admin on the License page. The plan's limits apply at once, until the expiry date shown. ### Prices Prices are on the [AmunCore home page](https://amuncore.com/). For a quote or an on-premises install, write to info@amuncore.com. Source: https://amuncore.com/docs/changelog.md ## AmunCore changelog > Every AmunCore release, newest first. The data API is versioned in its path (`/api/v1`); the releases below add to it. ### v3.59.1 (November 2026) - Arabic wording: a weekly answer or reminder now reads "كل يوم خميس الساعة 16:00" (it said "كل الخميس"). ### v3.59 (November 2026) - Repeating answers: "كل يوم الساعة 9 ابعتلي مبيعات امبارح", "every day at 9 send me yesterday's sales", "every Sunday at 10 send me last week's sales", or with the repeating words last ("ابعتلي مبيعات امبارح كل يوم الساعة 9"). A Yes first; then each day or week at that time the question is asked as the person and the fresh answer comes to the chat, for up to 90 days. "my alerts" shows the next time; "stop alert 3" stops it. - Repeating reminders: "remind me every day at 8 to take the medicine", "ذكرني كل خميس الساعة 5 اقفل الحسابات", also for someone else. - Help: a Top button on long pages; Topics on a phone opens at the top of the page. ### v3.58.1 (November 2026) - Invited people can remind each other: on an app for people chosen by email, the Channels page, Alerts tab, has Who can send to someone else: Only the team (the usual) or The team and people invited by email. Then they send reminders and alerts to each other and to the team, in their chats. - A name with the email ("فكّر نسرين nesreen@company.com كمان دقيقتين تتعشى"): the email decides who it is. - help in a chat also understands a few of its words together ("help مساعده"). - Help: the search stays at the top of the page while you read, and What people can write explains the help command (The help command in the chat). ### v3.58 (November 2026) - What people can write: a new Help page with every case in one place (connect a chat, ask, reminders, an answer at a time, alerts, for someone else, changes, approvals, and who can do what), in English and Arabic. In a chat (the app, Telegram, WhatsApp), help or مساعدة lists what that person can do there. - Reminders for someone else: "remind Ahmed in 2 hours to call Nile Foods", "ذكّر أحمد بعد ساعتين يتصل بنايل فودز". It goes to their chat at that time, with who set it, and the person who set it is told it arrived. - People by email: reminders and alerts for someone else find a person by their whole email too ("remind sara@company.com ...", "tell sara@company.com when ..."), including people invited by email, not only team users. - An answer at a time: "كمان دقيقتين هاتلي كل المواعيد", "in 10 minutes show me today's bookings", "tomorrow at 9 send me yesterday's sales". The question is asked at that time and the answer comes to the chat. - "add a reminder ..." in a chat with changes allowed is a reminder, not a new record. ### v3.57.2 (November 2026) - Reminders in any wording: "Reminder me in 2 minutes to call Nile Foods", "reminder: in 2 hours, call Omar", "can you remind me after 10 minutes...", "set a reminder for me in 3 days: ...", "call Nile Foods, remind me in 2h", "اعمل تذكير بعد ساعتين ...", "تذكير بعد 5 دقايق ...", "ممكن تفكرني بعد ساعة ...". What to remind is free text. A message that only mentions reminders without a time ("how many reminders were sent") stays a question. ### v3.57.1 (November 2026) - Chat invitations: on the Channels page, Invite tab, an admin emails each person (team, people chosen or clients) their own link to use the agent on Telegram and WhatsApp, without opening the app first. A button for the phone and a QR code for a computer; each link works once, for that person, for 48 hours. - QR codes: Use on Telegram and Use on WhatsApp in the app also show a QR code to scan with the phone. An app for anyone has a QR code to print or put on a website (Invite tab, Download the QR code). ### v3.57 (November 2026) - Reminders in chat: "remind me in 3 days to call Nile Foods", "ذكرني بكره الساعة 9 ابعت التقرير": a Yes first, then one message at that time. - Event triggers: alerts that go to someone else: a person who connected their chat ("tell Ahmed when there is a new booking"), or for each new record the person a column names ("when a new booking arrives, message the doctor"). ### v3.56.1 (November 2026) - Changes from chat with the built-in engine (no AI): "add a booking" asks for each required field, then shows the record with Yes and No; changes and deletes by the record's number. - Undo: a record added from a chat can be removed within 5 minutes (Undo, تراجع). - Cancel a request: a change sent for approval can be cancelled from the chat it was asked in (Cancel request 12, إلغاء الطلب 12). ### v3.56 (November 2026) - Changes from chat: a change asked in the app, the website chat, Telegram or WhatsApp is shown to the person exactly as it will be saved, with Yes and No, and saved only after their Yes (within 10 minutes). It goes through an endpoint that writes, with its rules, approvals and audit log. - Approvals in chat (Channels page, Approvals tab, off by default): approvers who connected their own chat get each waiting change with Approve and Reject; the person who asked hears the decision in their chat. ### v3.55.4 (November 2026) - AI Agents: AmunCore's own assistants are now called agents everywhere (pages, Help, docs, Telegram and WhatsApp replies, App Builder), in English and Arabic (وكيل). The offline brain is the built-in engine (no AI). Nothing else changes: settings, links, names you gave your agents and their history stay as they are. - "AI assistant" now means only outside tools such as Claude and ChatGPT that connect over MCP. ### v3.55.3 (November 2026) - App Builder in the plan and the license: apps (Free 1, Starter 5, Pro 50, Enterprise unlimited), who they may be shared with (Free: the team; Starter: also people chosen by email; Pro and Enterprise: also clients and anyone), email reports (Starter and above), alerts (Pro and above) and TV mode (Starter and above). A license key can set its own numbers. - Create with AI on every plan: with an Online AI agent, or a key pasted for that one request (used once, not saved). - When a plan changes, apps are kept: past the number no new one is made; an app shared in a way the plan no longer includes opens only for the team; reports, alerts and TV mode wait. - My apps and the License page say what the plan allows; the designer shows what it does not, with the plan that has it. ### v3.55.2 (November 2026) - Alerts on new records: "tell me when there is a new booking for Mohamed Samy", "قولي لما يكون في حجز باسم محمد سامي", "نبهني لما يتسجل طلب جديد". It writes when there is a new one, then shows the new records. - Alerts on a change, a rise or a fall, with no number: "tell me when the price of rice changes", "نبهني لما المخزون يقل". And "just once" / "مرة واحدة": the alert ends after its first message. - Channels page: Alerts, Telegram and WhatsApp lists in tabs; long lists in pages; How alerts work in a popup. - The built-in assistant reads "باسم" and "named" ("حجز باسم محمد سامي"), and booking / حجز / الحجوزات as appointments too; an answer that did not find the words never gives an alert a figure. ### v3.55.1 (November 2026) - Alerts in chat: in their own Telegram or WhatsApp chat with an assistant, people set alerts in words ("tell me when today's sales are below 50,000", "نبهني لما المخزون يقل عن 10"); it checks as that person, as often as asked (15 minutes at the most often, an hour with online AI), and writes only when the condition becomes true. "my alerts", "stop alert 2", "stop all alerts". - Each alert is set only after the person's yes: the assistant replies with the figure now and "Set this alert?", with Yes and No to tap (Telegram, WhatsApp and the app). - Who can set alerts: the assistant's team and invited people (the usual), or anyone who can chat with it. - Works with online AI too (asked for the figure only at each check, an hour at the most often). - Off until an admin switches them on (Channels page, Alerts in chat), where every alert shows with its figure, last check and last message, and can be stopped. - In the app, a person who connected a chat can ask for an alert there too; it comes to their chat. 10 alerts a chat, 300 a company; each ends after 90 days. ### v3.54 (November 2026) - Help Center redesigned: one search that answers questions (in English, and common Arabic words), with the best section first; Start here; topics as lists; What's new. - Every Help page in one frame: the topics on the left, Help / its area, the search, On this page on the right, Previous and Next; numbered steps, Copy on code, one look for every page; works on phones (Topics button). - Assistants dashboard: every long list shows in pages (10 rows each; 4 main assistants), with the page numbers; the latest questions keep their page when they update. Not answered, Who asks most and Asked most now show up to 50. ### v3.53.3 (November 2026) - Assistants dashboard (in the menu under AI access, and on each assistant's page): questions by channel (app, website chat, Telegram, WhatsApp, other systems, scheduled), the latest 50 questions (updated every 30 seconds), where each main assistant sent its questions and how many found no assistant, the questions not answered (with Teach), answer times (offline and online), who asks most and busy hours. - Filters by assistant and channel; the CSV follows them. - Each run keeps where the question came from and the main assistant that sent it; a question no assistant could take is kept on the main assistant. - A main assistant's answers can be rated in its app (kept on the assistant that answered). ### v3.53.2 (November 2026) - Assistant app refreshed: answers in cards, a total as a number card, a trend or top list with its chart first, Ask next under each answer, dark mode, and a language button (English or Arabic) for apps set to Automatic. - Main assistant's app: a card for each assistant with its questions, Ask to choose who answers, and each answer under the assistant's name. - Agent page: the Assistants step says which assistants a main assistant can use with the app chosen. - Charts of months by name are in the order of time. ### v3.53.1 (November 2026) - Main assistant is now its own kind, chosen under Basics: its steps are Basics, Brain, Assistants, App, Advanced and Review (no data, ready questions, steps or schedule of its own). - The Assistants step shows who each assistant is for, and marks the ones this main assistant cannot use with its app (for example a team assistant from an app for anyone) with how to fix it; saving is refused until fixed. A main assistant saved that way before now says why instead of "No assistant is available". - The same assistant in Arabic and English: the question's language chooses, instead of asking which. - The agent page's steps redesigned: a progress line, the current step marked, and one bar for the buttons (Back on the left; Cancel, Save and Next on the right). ### v3.53 (November 2026) - Main assistant: one assistant that sends each question to the assistant it is about (Sales, Clinics, HR...), with one app, one Telegram bot and one WhatsApp number for all. Follow-ups stay with the same assistant; when unclear it asks with buttons (or its AI model picks); a question about two assistants gets both answers. Each person reaches only the assistants their access allows. ### v3.52.2 (November 2026) - My mailboxes was removed (the page, the assistant setting and its tables, with the saved mailboxes and passwords, deleted at the next start). Scheduled results by email, Telegram and WhatsApp are unchanged. ### v3.52.1 (November 2026) - The Email channel (an assistant answering a mailbox by itself, with its card, sender checks and "Ask by email" in the app) was removed: Telegram and WhatsApp stay. Saved email channels and their messages are deleted at the next start. Scheduled results by email, and My mailboxes, are unchanged. ### v3.52 (November 2026) - My mailboxes: each person connects their own mailboxes and asks an assistant about them in the app, on Telegram and on WhatsApp: the new emails, what one says, and a reply (with their words, or with data from the records). A reply is sent only when they press Send. An administrator switches it on for the company; the assistant must have Each person's own mailboxes ticked. ### v3.51.2 (November 2026) - Email: AmunCore checks the sender's DKIM signature itself, so mail from Gmail, Outlook and most companies is answered even when your mail host adds no sender check. ### v3.51.1 (November 2026) - Scheduled reports in an app for anyone: the Channels page lists the people who wrote to the bot (Telegram and WhatsApp), each with a Reports switch. - Agent page: Save on every step; if something needed is missing, that part opens at the box to fix and nothing is saved. - Email card: Test (signs in without saving); Save tests first, so a mistake never loses what was typed. - Email sender check: the card shows the last refused email's checks; choose which mail server's check counts (mail through a spam filter). ### v3.51 (November 2026) - Email channel: people can email the assistant. A mailbox only for it (IMAP and SMTP, Gmail and Zoho with an app password, or your hosting's mail), read every minute; each new email answered in the same thread, to the sender only, with long lists attached as Excel or CSV and attached photos read. Who may email follows the app's access, by the sender's address; an email is answered only when the receiving mail server proved the sender (SPF, DKIM or DMARC). The app shows Ask by email. - Send now: in the Schedule, runs once and sends the result by Email, Telegram and WhatsApp as ticked, then says where it went (or why not). Each send is kept in the run's steps. ### v3.50.1 (November 2026) - Security: the files API (/api/v1/{app}/files) needs an API key (before, with no key the application was found by its id alone); a read-only key cannot upload or delete files, a write-only key cannot read them; the application's IP allow list applies to files and GraphQL too; the caller's address is the connection's own (an X-Forwarded-For written by the caller no longer passes the allow list or the rate limits; behind a proxy, Security:TrustForwardedFor uses the address the proxy added); password reset emails and the "is this email registered" check are limited. - Assistant app: Read photos (optional): a photo button in the message box; the text in the photo is read on the server and asked, with the same rules as Telegram and WhatsApp. Also in the chat button on your website. - Channels page: How to set up Telegram and How to set up WhatsApp open in a popup. ### v3.50 (November 2026) - The built-in assistant understands more, for any kind of data: "late" and "overdue" (not finished and past the due date), "open" and "not finished yet" when no status is called that, "finished", "due today", "due tomorrow", "upcoming", "in the next 7 days", "next week" and "next month" on the due date, "who has the most late jobs", "which technician has the most open jobs", averages for grades and scores ("best student"), and many more Egyptian Arabic phrasings (الشغل المتأخر، معاده النهارده، فات معاده، المهام اللي لسه متخلصتش، المواعيد الجاية، مين أشطر طالب، Delta Trading عليها كام). Checked against a bank of 183 questions over sales, field service, clinics, schools and projects. - Channels page: the messages are one short line each; press one to read it whole. 15 to a page, newest first. - AI access page: a clearer layout (the switch, the figures, each connection and agent with its facts), on phones too. - Security: the diagnostics page (/Diag) is for a SuperAdmin only and no longer shows the database password; the emergency 2FA reset needs a key of 16 characters or more and allows 5 tries an hour; the assistant app pages send a strict Content-Security-Policy; the encryption keys can be locked with a secret kept outside the database (Security:KeyEncryptionSecret, opt-in). ### v3.49 (November 2026) - Business pulse: ask "pulse", "what needs my attention" or "نبض الشغل" (or set it as the question on each scheduled run) and the built-in assistant shows, over every list and for any kind of data, what is late and who has it, what is due this week, who has the most open items, what changed against last week and what is unusual, with the figures and a table of the most late. - New template: Business pulse every morning. ### v3.48 (November 2026) - Send each run's result by Email, Telegram and WhatsApp: three ticks in the assistant's Schedule. Who gets it is still chosen on the Channels page (Reports On), now also for WhatsApp chats. - WhatsApp reports use a message template Meta approved (WhatsApp lets a business write first only that way): give its name and language in the Schedule; the result fills its one variable, on one line. ### v3.47.1 (November 2026) - Photos, clearer: icons and borders that the reader turned into symbols are left out of what was read. The offline assistant answers a photo's caption (the read text is shown, not searched); a photo without a caption shows what was read and asks what you want to know. Assistants with an AI model get the read text with the caption, or are asked what the photo shows. - "I could not find anything about ..." names at most 6 words. ### v3.47 (November 2026) - WhatsApp setup: the steps on the Channels page and in Help follow Meta's current flow (Create App with the use case Connect with customers through WhatsApp, API Setup, a system user's token, Configuration, publish). Saving subscribes the app to the WhatsApp account, so real messages reach AmunCore. - Questions to tap: an answer's suggested next questions and "did you mean" choices come as buttons on Telegram (keyboard) and WhatsApp (reply buttons or a list). - Telegram groups: switch on "Also answer in groups I approve"; people ask with /ask or by replying to the bot; each group waits on the Channels page for an admin to approve, stop or remove it; only people who connected their own chat can ask, each with their own access; the answer replies to the question (in its topic). - Reports to Telegram: a connected chat or an approved group set to On under Reports gets each scheduled run's result. - Photos: a photo's text is read on the server (OCR, English and Arabic) and asked with its caption. - Voice notes: assistants with an AI model hear them with their own provider (OpenAI, Groq, Gemini, or a Whisper server); the offline assistant asks to type. ### v3.46.1 (November 2026) - The encryption keys are kept in AmunCore's database (table DataProtectionKeys, made on first start), so a restart or an update no longer loses them. Before, on hosts with no user profile (IIS, shared hosting), every restart lost the saved Telegram and WhatsApp tokens, an assistant's own AI key and sign-ins. After installing, paste the Telegram or WhatsApp token once more; people sign in once more. - Channels: when a saved token can no longer be read, the page says so (Check, Save, and Last problem when messages arrive) instead of Telegram's "Not Found". ### v3.46 (November 2026) - Telegram and WhatsApp: a chart comes as a picture (a line for months and dates, bars for the rest; Arabic right to left), and a long list comes as an Excel file (right to left in Arabic, numbers you can add up) or a CSV file, chosen per channel ("Long lists also come as"). Pictures in the records (a product's photo, by its link or kept in the database) come as pictures with the record's name, up to 6 an answer. - Pictures in records: the app shows picture links and pictures kept in the database (binary columns) as pictures in lists and details, and pictures written in an AI answer. Stored pictures are made small for the answer; other stored files show "(file)". The AI model is sent "(picture)" or "(file)", never the bytes. Also in App Builder apps. ### v3.45.1 (November 2026) - Telegram and WhatsApp: a CSV file with an Arabic name (or any non-English name) now shows that name in the chat. Telegram showed it as a code like "=?utf-8?B?...?=" before. ### v3.45 (November 2026) - Team questions: the agent's ready questions are also in its app's side panel, by group, at any point of a chat. A company admin using the app adds a question under any answer (Add to team questions), into a group they pick or a new one. - Telegram and WhatsApp: a chat app's agent can also answer on Telegram (free, a bot from @BotFather) and WhatsApp (Meta's WhatsApp Business Platform), from the agent's new Channels page. The same answers, people and rules as its app: anyone for an app for anyone; otherwise each person connects their chat once from the app with a one-time code, checked again at every message. Long lists also come as a CSV file. Connected chats and the latest messages are on the Channels page. ### v3.44 (November 2026) - Dark theme: the saved theme is set before the page is drawn, so pages no longer show light for a moment first (the menu pages, sign in, register, forgot password and the website pages). The logo follows at once. - AI access, agents: a published app has Open app next to Copy, opening it in a new tab (while the agent is on). ### v3.43 (November 2026) - Recently deleted: a deleted application can be restored for 7 days, with the assistants, Client Connectors (and their clients) and App Builder apps deleted with it. Before restoring, every name and address is checked against what exists now; anything used now gets a free one, shown first. Refused when the plan has no room for all of it. Delete for good removes one at once. - AI assistants: saved and recent questions belong to the assistant itself. A new assistant given a web address used before (for example the sample data made again) no longer shows the old one's questions. - Loaders on Test connection, App Builder Publish and Restore. ### v3.42 (November 2026) - Deleting an application now also finds: assistants that read it only in their steps (an endpoint, or one of its data files), App Builder apps that show its documents or data files, App Builder apps whose Ask block talks to an assistant that is deleted, and what an earlier delete left behind (marked left over). It uses the application's own company, so a SuperAdmin's delete finds them too. - Deleting an application and adding sample data show a loader with the steps while the work is done. ### v3.41 (November 2026) - Deleting an application now shows, by name, everything that goes with it (endpoints, the assistants and agents that use it, the Client Connectors that share it, the App Builder apps that show its data, live feed links), asks for its name to be typed, and deletes all of it. Before, only its endpoints went and the rest was left broken. - App Builder: column names (Theme): the name shown for each column in table headings, figure labels, titles made from columns, exports and email reports. The Arabic sample apps have Arabic column names. - App Builder: an app whose words are Arabic sends its email reports and alerts in Arabic. - API health: the people to tell are chosen from a searchable list (name and email, up to 20). - Dashboard: the first step offers two clear ways to start: connect your database, or explore with sample data. ### v3.40 (November 2026) - App Builder: Words on buttons (Theme): Arabic puts a published app's own words in Arabic (Refresh, More, Notes, filters and their periods, "Updated", the notes under tables, chart switches, the files list, the Ask block). Titles, column names and data stay as written. The Arabic sample apps have it on. - Live feed for Excel and Power BI: an endpoint that reads (GET) can have a secret link (Endpoints, Feed), as CSV for Excel or JSON for Power BI, refreshed whenever they like with no key. It reads with the access of the person who made it, up to 20,000 records, 240 reads an hour; shown once, replaced or switched off at any time, only its hash kept. CSV text that looks like a formula is written so Excel never runs it. - API health (menu, admins and developers): each endpoint over the last hour, day or week (calls, failed, refused, average and slowest answer, last failure; Fine, Failing, Slow, Not called). Email alerts, off until an admin turns them on: checked every 15 minutes, one email when an endpoint starts failing or slowing down and one when it is back to normal. ### v3.39 (November 2026) - Try with sample data: each trade added also gets an App Builder app, published for the team: pages made from the App Builder templates on its sample tables (for Retail: Sales, Products, Collections, Customers) and an Ask page with its assistant. Arabic sets get the app in Arabic, right to left, with the records' own Arabic words. One click gives the data, its API, an assistant and dashboards. ### v3.38.2 (November 2026) - Demo Database: the older two-table demo is removed (its Table 1 and Table 2 fields, and the Oracle choice, which sample data cannot use). Saving the page clears the two saved table names; the Dashboard offers Try with sample data only once sample data can be built. ### v3.38.1 (November 2026) - Try with sample data: a sample application made earlier (the older two-table demo, or before the sample data moved to another database) now follows the current sample data the next time the company adds a trade. Its connection is updated and the two older demo endpoints are removed; endpoints the company added itself are kept. The older demo tables can then be deleted. ### v3.38 (November 2026) - Try with sample data: 8 trades (Retail and POS, Distribution, Clinics, Real estate, Restaurants, HR and payroll, Schools, Manufacturing), each in English or Arabic, with 18 months of linked records. A company picks one or more and gets read-only endpoints and a ready assistant for each, within its plan. The platform administrator builds them with one button (Demo Database); their dates move forward by themselves. - A new Dashboard: your numbers (applications, endpoints, assistants, API calls, users), what you can do next, API calls by day for 14 days, recent requests, your assistant apps, and the first steps for a new company. - Built-in assistant: "by branch" uses the Branch name, not BranchId, when a list has both; "الشهر الماضي" is never read as a mistyped list name. ### v3.37 (November 2026) - Built-in assistant: each answer says its figure once ("العدد: 20 من Customers." no longer comes with a "Customers: 20" box under it); a figure the sentence does not say still shows; reports keep their tiles. ### v3.36 (November 2026) - Agent apps: Ask by voice and Read answers aloud are paused. The browsers' own speech was not good enough, above all in Arabic; they return with AmunCore's own speech engine. Saved choices are kept. ### v3.35 (November 2026) - Agent apps: Short answer for the built-in assistant: the answer in words with up to 4 figure cards, and each list as a card with its count, its first three and Excel, Word and PDF files (made in the browser, Arabic right to left). People can switch to the full report. - Read answers aloud: a Listen button under each answer (Arabic or English). ### v3.34 (November 2026) - Agent apps on your website: one line pasted on your pages adds a chat button, only on the websites you list. An app for anyone opens in a panel on the page (full screen on phones); an app that signs in opens in a new tab. - Ask by voice: a microphone button in agent apps (also in the panel); the words spoken appear in the box, in Arabic or English, with the matching questions. ### v3.33 (November 2026) - App Builder: on the last step of a template's setup only Back and Make the app show (Next stayed and led nowhere); on New app, A template always goes to the templates (it could make a blank app from the step chips), and Make it works once. - Agent apps: the question board opens folded (one row and All questions) so the chat stays in view; questions it answers are offered while typing. - Ready questions: the questions people ask most come first; 28 trades (added finance and lending, insurance, agriculture, car rental, travel, facilities, salons, events, projects and agencies, charities, medical labs, law firms). - Built-in assistant: more Gulf wording (ما انسددت, بكل مدينة, الشهر اللي طاف, بالمبيعات, جيب لي, منو لازم اتصل فيه, مقارنة بـ, الفلوس اللي برا عند العملاء). ### v3.32 (November 2026) - Agents: apps can be opened by people you choose by email, or by the invited clients of a Client Connector, who sign in with a 6-digit email code (read only; clients read as themselves, with the connector's rules). - The coverage score: the share of people's questions each agent answered, this week against last, with 8 weeks of bars (Activity, the Agents dashboard and the weekly email). - Ready questions in Arabic (answered in Arabic), and 16 trade packs: added Restaurants and cafes, Manufacturing, Schools and training, Pharmacies, Hotels and stays, Car service and workshops, Construction and contracting, Logistics and delivery, E-commerce, Gyms and clubs, People and HR, Customer service. - Built-in assistant: "top 5 invoices by Amount" lists the 5 largest (it grouped by the amount before). - App Builder: New app and template setup are step by step (or all on one page), with a review before the app is made. ### v3.31 (November 2026) - Agents: a new agent is made step by step (each part checked: a name, the person it works as, what it can use, a step's endpoint), ending with a review; or all on one page, with what is missing marked on the menu. - The built-in assistant needs no steps: with applications or endpoints ticked under What it can use, it reads the lists a question is about. Steps use only what is ticked there (refused otherwise). - Ready questions are made from the page as it is, also for a new agent before saving (the button did nothing on a new agent before); every list the agent can use gets its own; new kinds: this month against last month, anything unusual. - Industry packs: Retail, Distribution, Clinics, Real estate add their questions and words where the data has the lists and columns they need. - Apps: two kinds, Questions and chat (ready questions as buttons, plus typing) and Form (look up a record or send a request), each shown with a picture. The Questions app of v3.30 is now Questions and chat. Template "Ask your data (no AI)". ### v3.30 (November 2026) - Agents: a clearer agent page: a side menu (Basics, Brain, Data and permissions, Steps, Questions, Words and learning, App, Schedule, Advanced), Save always in view, and at the top Try it, Activity (What people said and Runs as two tabs) and More (Run its job now, Copy, Export). The New agent page groups templates by what they do, with a No AI and Online AI filter. - Ready questions replace check questions: made by AmunCore from the agent's data, added to or removed by people, and tested in one click with a score ("Answers 18 of 19"). - A new app type, Questions: the ready questions as buttons grouped by list, with Period and value lists (customer, city...) to narrow them; the answer shows as a report. New template: Questions app (no AI). ### v3.29 (November 2026) - Built-in assistant: teach a whole question (built from menus in the test console, checked, then taught); it is answered as what it means, also with other customers, cities, numbers or periods. "Did you mean" offers up to 3 questions when it cannot answer; with "Learn from what people choose" ticked, the one pressed is remembered. What people said lists the questions it could not answer and those marked wrong, ready to teach, and the taught ones; the weekly summary lists the questions not answered. ### v3.28.1 (November 2026) - Agents: Excel, CSV, JSON and XML files are read through Documents: tick Documents for their application and the built-in assistant reads them as records (each sheet of a workbook its own list). Get records reads endpoints only again (the file choice added in v3.27 is removed). ### v3.28 (November 2026) - App Builder designer: a clearer Tools panel (blocks as tiles with pictures, grouped as Data and Content; pages, filters and app settings with pictures, the page in use marked, its number of blocks shown) and Properties panel (what is chosen at the top with its picture, larger fields with a focus ring, sections easier to tell apart). - Built-in assistant: an ageing report ("ageing report", "أعمار الديون"), a collection list ("who should I call first", "قائمة التحصيل": the costliest first, with the oldest invoices, the phone and email, and what to do), and a cash forecast ("cash forecast for the next 13 weeks", "توقع السيولة": cash in and out week by week, with each customer's usual delay, and the balance each week when today's cash is given). - App Builder: two finance templates: Ageing and collections, and Cash forecast. Each has figures and charts and reports that answer when the page opens. ### v3.27 (November 2026) - Built-in assistant: an agent can read a data file (Excel, CSV, JSON or XML uploaded under Files) as a list, in Get records; a built-in assistant in an app also reads the files its blocks show. A customer kept as a number (CustomerID) is named from the customers list; amounts in several currencies are answered one currency at a time ("... in AED"); an invoice's total is its amount, not its tax or discount. ### v3.26 (November 2026) - App Builder templates: optional lists are marked Optional and say when they are needed; notes appear only when they help (a column of the wrong kind, a column the app needs most left out, a list that does not look right); the data is suggested by its columns as well as its name. - Built-in assistant: on the first day of a month, "this month" is set next to the same day of the month before (not yesterday); "analysis of invoices this month" with only a few invoices is still an analysis; a list question that finds nothing on a long list says it was checked by the database over every record. ### v3.25 (November 2026) - Answers on a phone: charts are drawn at the phone's width, so their words keep their size; in wide tables the first column (the name) stays in view while the others scroll. - App Builder: a report that answers when the page opens grows to fit its answer, and the blocks under it move down (no scrolling inside it). ### v3.24 (October 2026) - Built-in assistant: the cost of late payment, profit by product or customer, and working capital are worked out by the database over every record of a long list (before: the first 10,000). - Agents: a Yearly cost of money (%) for the cost of late payment; a rate in the question wins. ### v3.23 (October 2026) - App Builder: three finance templates: Late payers and what they cost, Profit by product and customer, Working capital and cash. Each has figures and charts and a report the built-in assistant writes when the page opens. - App Builder: an Assistant block can answer its first question when the page opens, like a report that stays up to date, and is printed with the page (built-in assistant only, so no AI is paid for on each visit). A built-in assistant in an app reads the app's own lists. - App Builder: conditions on a date can be said from today (today, today-30, today+7). - The version (v3.23) is shown in Settings and in the platform settings. ### v3.22 (October 2026) - Built-in assistant: the questions a finance manager asks. "What are late payers costing me" (what is owed past the due date by customer, and its cost at a yearly rate); "what should we sell more of" and "which clients am I losing money on" (contribution by product or customer, with Grow, Review, Reprice, Investigate or Data missing); "what is shrinking working capital" and "why is there never cash" (DSO, DIO, DPO, the cash cycle, and where cash is stuck). In English and Arabic; what is missing is named, never guessed. ### v3.21 (October 2026) - Built-in assistant: an analysis of a list ("I need details with analysis for invoices", "حلل الفواتير"): what stands out, the figures, the trend by month, the top 5 and their share, what is owed and late, the largest records, and how it worked it out. A question that names a list is answered from it, not from a document that shares its words. - Answers look like a report: figures as tiles, titles, findings with markers; App Builder assistants have a clearer header, a new conversation button, a working indicator and a copy button. ### v3.20.1 (October 2026) - App Builder sign-in page: with the layout "A card in the middle", the sign-in picture now shows across the top of the card (before, only the other two layouts showed it). - Home page: an example dashboard in English and Arabic, clearer pricing and questions sections, balanced headings. ### v3.20 (October 2026) - App Builder designer: dragging with the mouse lands exactly where the marker shows: the page stays still while you drag, and over another block you choose beside, above or below it; the rows a block leaves close up. - App Builder designer: arrange several chosen blocks (one row, spread evenly, fill the row, same width, same height, align tops), and choose blocks by dragging a box around them. - App Builder charts: area, pie, stacked bars (split by a second column) and bars with a line (two values); a figure's target as a gauge. - App Builder printing and PDF: blocks keep their widths and rows on paper, each as tall as it needs; long tables run over pages; landscape or portrait. - App Builder right to left: number columns line up with their headings, notes sit on the right, page arrows point the Arabic way. - App Builder: a title with a {Column} leaves it out when there are no records, instead of showing the braces; big pictures (sign-in, logo, blocks) and connector logos are made smaller in the browser to fit. ### v3.19 (October 2026) - App Builder designer: the tools in sections that fold; blocks dropped beside another share its row; pressed blocks line up; faster scrolling while dragging; a button to move a block to the top; smoother dragging that never drops out. - App Builder motion: a figure that changes when the page is read again glows once; pages slide in and the latest point of a line breathes (Lively). ### v3.18 (October 2026) - App Builder: right to left (Arabic) for a whole app or one block, with tables, charts, email reports and alerts. - App Builder: block titles with a column's value ("Balance of {AccountName}"); conditions and color rules with several values separated by commas. - App Builder: the designer's properties in sections that fold. - Built-in assistant: understands the columns and values of many more kinds of business in Arabic and English, with no words taught. - Data API totals: conditions for any of several numbers, none of them, and text that has any of several words. ### v3.17 (October 2026) - App Builder: four accounting templates from journal lines: trial balance, account statement (ledger), income statement and balance sheet, and journal analysis. - App Builder: pivots with several values side by side, each over the dates chosen, before them (opening) or up to their end (closing); a second column next to the first. - App Builder: figures that are opening or closing balances; charts of a formula; tables with a running balance for each account. ### v3.16.1 (October 2026) - App Builder: alerts and email reports can go to email addresses you type, for people who need the information without access to the app (an admin sets this). - App Builder: email reports can go to everyone on the email list of an app for people chosen by email. ### v3.16 (October 2026) - Data API: **dry run** for writes (`?dryRun=true` or `X-Dry-Run: true`): every check runs, a change or delete counts the records it would touch, and nothing is written. - Data API: `X-RateLimit-Limit` and `X-RateLimit-Remaining` on every answer of a plan with a call allowance; `Retry-After` on every 429. - Data API: a written OpenAPI 3.0 description with examples, code samples, idempotency, dry run and reversibility for every operation. - Discovery: `/.well-known/api-catalog` (RFC 9727), `/apis.json`, `/status`, and AI usage signals in `robots.txt`. - App Builder: alerts can be emailed to everyone on the email list of an app for people chosen by email. ### v3.15 (October 2026) - App Builder: alerts when a figure passes a line, and notes on pages and figures. - Built-in assistant: a question answered and emailed on each scheduled run (a morning briefing); averages and groups compared with the same days of the period before. - Plans: Starter includes MySQL as well as SQL Server. ### v3.14.1 (October 2026) - `llms.txt`, `llms-full.txt`, public Markdown documentation, `robots.txt` and `sitemap.xml`. ### v3.14 (October 2026) - AI access and Client Connectors: `get_totals` and `find_records` tools, worked out by the database over every record (off until switched on). - App Builder: search looked for by the database, and motion (Off, Subtle, Lively). ### v3.13 (October 2026) - Built-in assistant: several questions at once, the period before, a briefing, the true first records chosen by the database, and more spoken Arabic. - External APIs that page their own way are read to their last page. ### v3.12 (October 2026) - App Builder: figures, charts and pivots worked out by the database over every record; apps for invited clients. ### Earlier in 2026 - v3.0 to v3.11: AmunCore Studio and App Builder (dashboards, report pages, data pages, templates, email reports, data files, apps for people chosen by email). - Agent Studio v1 to v2.4: AI agents with your own key and the built-in assistant, schedules, and agents published as apps. - Client Connectors, who sees which rows and columns, approval policies, Shapes with XML and frozen copies, and AI access over MCP with OAuth.