Docs for your AI.

An MCP server that gives your AI assistant fast, ranked documentation search across Rust, Python, and JavaScript. Open source.

.mcp.json

{
  "mcpServers": {
    "docsnippet": {
      "url": "https://cloud.docsnippet.com/mcp"
    }
  }
}

Or install with one click:

search_docs

reqwest

Three ecosystems

Crawls docs.rs, npm, and PyPI. Pass a library name — the ecosystem is detected and indexed automatically.

Rust JS Python

Cache, BM25, cross-encoder reranking, token budgeting, session deduplication. Every query passes through all five.

Results are truncated to fit your LLM's context window. No wasted tokens on irrelevant content.

Works with Claude, Cursor, Windsurf, and any MCP client. One config line to install.

Every search passes through a five-stage cascade that narrows thousands of pages down to the most relevant snippets.

1Cache

Instant hit for repeated queries

2BM25

Full-text keyword matching

3Cross-encoder

Neural reranking for relevance

4Token budget

Trimmed to fit context window

5Session dedup

No repeated results across turns

ecosystems

search layers

—

libraries

1 line

to install

Free up to 100 requests/day. Pro for $8/mo.

Unlimited requests, deep search, API keys, and priority indexing.