Skip to content

v0.2.0 — reliable MCP web search and tiered fetch

Latest
Latest

Choose a tag to compare

View all tags
@Dvorf Dvorf released this 01 Sep 16:25
· 1 commit to main since this release
v0.2.0
adfed87

web-retrieval-mcp v0.2.0 is the first self-contained public release of the current retrieval stack: six read-only MCP tools for source-grounded web search, resilient page fetch, AI and ML paper discovery, and developer-source research.

Highlights

  • Select Exa or Tavily for web search, with caller-visible provider routing and Firecrawl fallback provenance.
  • Fetch through an adaptive Exa → guarded Camoufox → optional Tavily Extract → Firecrawl cascade.
  • Request full page text, concise summaries, or grounded answers with freshness and size controls.
  • Configure Exa, Firecrawl, and Tavily credentials through environment variables, a private key file, keyring, macOS Keychain, or Linux Secret Service.
  • Install a lean base package or opt into rendering, Tavily, Valkey caching, keyring, or the complete feature set.
  • Run over stdio or loopback Streamable HTTP.
  • Connect from Codex CLI and the Codex IDE extension, Claude Code, Claude Desktop, Cursor, ChatGPT desktop, or another compatible MCP client using a transport supported by that client.
  • Keep explicit source, cache, truncation, and routing disclosures plus fail-closed initial URL and browser redirect SSRF checks.

Tavily contributions

Tavily Search PR #7 and Tavily Extract PR #8 were opened against the former v0.1.2 tree. Version 0.2.0 already contained independently implemented, stricter versions of both capabilities, so the PRs were transparently merged with an ours conflict resolution: their commits remain in history for attribution, while their older conflicting file trees did not replace the reviewed 0.2.0 snapshot. Thank you to @tavily-integrations for proposing both integrations.

Install from this GitHub release

Python 3.10 or newer is required.

python -m pip install "web-retrieval-mcp[all] @ https://github.com/VelvetSP/web-retrieval-mcp/releases/download/v0.2.0/web_retrieval_mcp-0.2.0-py3-none-any.whl"
python -m camoufox fetch

Then configure one or more provider credentials:

export EXA_API_KEY="<your-exa-api-key>"
export FIRECRAWL_API_KEY="<your-firecrawl-api-key>"
export TAVILY_API_KEY="<your-tavily-api-key>"

This operation publishes a GitHub release and attached package artifacts only. It does not publish v0.2.0 to PyPI or the MCP Registry.

Verification

  • Release commit: adfed87157a3e011141ad42447d23698a473237f
  • Reviewed tree: b67ed2baf1686f523b83577c504c8b620d044901
  • GitHub Actions CI: Python 3.10, 3.11, 3.12, and 3.13 base installs plus the complete public release gate passed.
  • Local ./run-tests.sh, wheel and sdist metadata validation, clean installed-wheel acceptance, secret scan, and package inventory all passed.

SHA-256

34cb36ef6c1adcdefc74a4ecde6e459f84a33633bc61c6474608151408be9bca  web_retrieval_mcp-0.2.0-py3-none-any.whl
42d81fd214268da2f385c171ab55c0265cea815c959ef6c94dc3a6b6a8f6e448  web_retrieval_mcp-0.2.0.tar.gz

See the README for client setup, routing, configuration, security, caching, and examples.

Assets 5
Loading