Quickstart

Get factq running in your project in under a minute.

1. Initialize

Point factq at your project root. It auto-indexes your docs/ folder:

factq init

This scans docs/, docs/research/, docs/knowledge/, and any markdown files alongside your code.

2. Query

Ask a question. factq checks your project docs first:

factq query "What is our API rate limit?"

If the answer exists in docs/architecture.md, it returns immediately — no LLM call, no network request.

3. Save Findings

When factq discovers a new verified fact through the Deepthink loop, save it to your docs:

factq save docs/research/rate_limits.md

The finding is now a markdown file in git — shared with your team on the next push.

Python API

import factq

fq = factq.init(".")

# Docs-first resolution
result = fq.query("What database do we use?")
print(result.answer)
print(result.source)  # 'docs' | 'kb' | 'deepthink'

# Save a verified result
fq.save("docs/research/database_choice.md", result)

As an MCP Server

factq runs as a local MCP server for Claude Code, Cursor, and Windsurf:

{
  "mcpServers": {
    "factq": {
      "command": "factq",
      "args": ["mcp"]
    }
  }
}

Your AI assistant now checks your project docs before every answer.