Comments
>log in to comment
Hi everyone 👋 We built FactQL because every "AI on your data" demo we tried gave an answer but no way to check it. Every FactQL answer comes with the query that ran, the rows behind it and how it got there. • Ask in plain English: Postgres, MySQL, DuckDB, CSV, Excel, PDFs • Answers as text, tables or charts, plus recommendations • Pin charts to boards; the query comes with them • REST API: one POST returns answer, query, rows and lineage • One API key for many end users, ask-only keys for browser apps, drop-in chat component • Open-source MCP server for Claude, Claude Code, Cursor and VS Code When the data can't answer, it says so instead of guessing. Free, no card, about five minutes: upload a CSV and ask. What confused you in the first five minutes, and which data source should we add next?
FactQL AI is an AI data analysis API and open-source MCP server. Add "ask your data" to any app or AI agent with one API call.
Send a question in plain English. You get back the answer, the query that ran, the rows behind it and how it got there, so every number can be checked. FactQL runs the query before it answers, and it can return charts and recommendations.
Data sources: PostgreSQL, MySQL, DuckDB, CSV, Excel (XLSX), PDF.
- REST API: one POST returns JSON with the answer, query, rows and lineage. One key can serve many end users with private conversations, plus a drop-in chat component for web pages.
- MCP server (MIT): connect Claude, Claude Code, Cursor or VS Code to your database.
uvx factql-mcpor the hosted endpoint. Listed in the official MCP Registry. - Web app: upload a spreadsheet or connect a database and start asking.
When the data can't answer a question, FactQL says so instead of guessing. Free tier, no card.
More AI tools →80 free dev tools →
Looking for MCP servers to install? MCP servers list: 289 servers by category, official and community →
FactQL AI provides an API that lets apps ask natural-language questions over databases, spreadsheets or PDFs and returns verified SQL, results and lineage.
- for
- Developers building AI agents or apps that need trustworthy data answers.
- pricing
- freemium · free trial
- license
- MIT
Key features
- One-call API — Send a plain-English question via a single POST and receive answer, SQL, rows and lineage in JSON.
- Execution-verified answers — The model drafts a query, runs it, verifies results and self-corrects before responding.
- Multi-source support — Connect to Postgres, MySQL, DuckDB, CSV/XLSX files or cited PDFs through the same endpoint.
- Fail-closed isolation — If verification fails the API returns a 4xx error, never guessing or exposing wrong numbers.
- Embeddable chat component — Add a drop-in chat UI for end-users with private conversations via X-End-User-Id.
- MCP server integration — Expose the service to Claude, Cursor, VS Code or any MCP-compatible app without installing anything.
Use cases
- Add a “ask your data” chat widget to a SaaS dashboard for instant KPI queries.
- Enable AI agents (Claude, Cursor) to retrieve and verify data from a company’s warehouse.
- Allow non-technical staff to upload a spreadsheet and ask ad-hoc business questions.
FactQL AI vs alternatives
 FactQL AI |  LlamaIndex |  LangChain |  n8n |  Apache Superset | |
|---|---|---|---|---|---|
| Best for | Verified natural-language data queries | LLM data-app framework | LLM workflow orchestration | Workflow automation | Data exploration & visualization |
| Pricing | Freemium | Free | Free | Subscription | Free |
| DevHunt upvotes | 1 | 0 | 0 | 0 | 0 |
| Launched | Dec 2026 | — | — | — | — |
- FactQL AI vs LlamaIndex: LlamaIndex is a framework for building LLM apps over data, while FactQL offers a ready-made verified query API.
- FactQL AI vs LangChain: LangChain provides generic LLM chaining tools; FactQL focuses on verified SQL generation and execution.
- FactQL AI vs n8n: n8n automates workflows, but does not provide built-in execution-verified data querying like FactQL.
- FactQL AI vs Apache Superset: Apache Superset visualizes data after it’s queried; FactQL generates and verifies the queries themselves.
FactQL AI FAQ
What happens if the generated query is wrong?+
FactQL runs the query first; if verification fails it returns a 4xx error and does not guess.
Is my data safe?+
Only SELECT queries are allowed, credentials are encrypted, and tenant isolation with committed proofs protects each user.
Can I start without giving production database credentials?+
Yes, you can upload a CSV or spreadsheet and query it before connecting to a live database.
How does pricing work when limits are reached?+
Questions are counted per plan; exceeding the limit returns a 402 error, so you never incur unexpected charges.
Do I need to install anything to use FactQL with my IDE?+
No, you just point Claude, Cursor, VS Code or any MCP-compatible app to the FactQL MCP endpoint and sign in.
Summarized by DevHunt from factql.ai · Nov 15, 2026. Details may change; check the official site.
