Skip to content

v1.0.0

Choose a tag to compare

View all tags
@KazKozDev KazKozDev released this 30 Jun 14:11
Immutable release. Only release title and notes can be modified.
v1.0.0
3f85f61

First public release of footnote-mcp — an MCP server for source-grounded web research.

It searches the web, fetches and extracts pages, pulls structured data out of tables/files/APIs, and — the part that sets it apart — verifies that a claim is actually supported by its source instead of trusting a snippet. Design priority: trustworthiness over convenience.

Highlights

  • 42 tools over stdio MCP — discovery/reading, structured data, source quality & verification, sandboxed extraction recipes, and a Chromium browser fallback.
  • Claim verification — evidence_entailment, corroborate_claim, and locate_claim_span check a claim against the source text. On numeric/factual data claims the offline heuristic catches 100% of unsupported claims (see benchmarks/REPORT.md); a local LLM judge (backend="ollama") closes the semantic gap.
  • Works with zero config — no API keys required: scraped Bing + DuckDuckGo search, an automatic headless-browser fallback, and an offline verification heuristic. Keys and env vars only make it better.
  • Provenance by default — every fetched page is cached with provenance under ~/.footnote-mcp/.

Install

pip install -e .
python -m playwright install chromium
footnote-mcp

Requires Python ≥ 3.10. See the README for MCP client config, search backends, and the fetch ladder.

License

MIT

Full Changelog: https://github.com/KazKozDev/footnote-mcp/commits/v1.0.0

Assets 3
Loading