pitlane-mcp
Token-efficient code intelligence MCP server for AI coding agents.
pitlane-mcp indexes a codebase once with tree-sitter and lets agents retrieve the symbols they actually need instead of dumping whole files into context.
接入你的智能体
從發布者安裝。不是 PlugMCP 端點。
遠端 URL 還是本機指令? 兩種安裝方式的差異
來自發布者
Token-efficient code intelligence MCP server for AI coding agents.
pitlane-mcp indexes a codebase once with tree-sitter and lets agents retrieve the symbols they actually need instead of dumping whole files into context.
Token-efficient navigation substrate for AI coding agents. Index code once and retrieve only the symbols you need.
pitlane-mcp is feature-complete and currently in maintenance mode. New major features are not actively planned. Bug fixes, compatibility updates, documentation improvements, and focused contributions are still welcome.
AI coding agents tend to over-read. pitlane-mcp gives them a smaller navigation surface and symbol-level reads so they can:
On the benchmark corpora in this repo, that translates into large reductions in prompt footprint, often by one or two orders of magnitude, while keeping navigation grounded in the codebase.
Download a pre-built binary from GitHub Releases for Linux, macOS, and Windows.
Or install via Homebrew:
brew tap eresende/pitlane-mcp
brew install pitlane-mcp
Or install via cargo-binstall:
cargo binstall pitlane-mcp
Or install from crates.io:
cargo install pitlane-mcp
Or build from source:
cargo build --release
cp target/release/pitlane-mcp ~/.local/bin/
cp target/release/pitlane ~/.local/bin/
claude mcp add pitlane-mcp -- pitlane-mcp
Add to opencode.json or opencode.jsonc:
{
"mcp": {
"pitlane-mcp": {
"type": "local",
"command": ["pitlane-mcp"]
}
}
}
Add to .vscode/mcp.json:
Optional semantic-search variables go under env.
{
"mcp": {
"servers": {
"pitlane-mcp": {
"type": "stdio",
"command": "pitlane-mcp",
"args": [],
"env": {}
}
}
}
}
Add to .cursor/mcp.json or .kiro/settings/mcp.json:
Optional semantic-search variables go under env.
{
"mcpServers": {
"pitlane-mcp": {
"type": "stdio",
"command": "pitlane-mcp",
"args": [],
"env": {}
}
}
}
Most users should stay on the default tool tier:
ensure_project_ready once at startup. It ensures the index exists and reports whether embeddings are still running, but it does not block on embeddings.investigate first for broad code questions such as subsystem, behavior, and execution-path questions.locate_code when you need discovery without full source.read_code_unit once you know the target.trace_path for explicit source-to-sink or config-to-effect questions.analyze_impact before edits or refactors, and analyze_changes to review a Git diff.search_content only when you know a text fragment but not the owning symbol.search_knowledge for documentation questions — it searches indexed Markdown (docs, READMEs, runbooks) by heading and content. It understands OKF v0.2 documents natively (okf_type/status/min_trust filters; results carry trust tier, staleness, and related-document links).read_knowledge_document to open a document found by search_knowledge — the full source Markdown by default, or one section via section. Results include OKF metadata, the section outline with line ranges, and related_docs/referenced_by links.Default public tier:
ensure_project_readyinvestigatelocate_coderead_code_unittrace_pathanalyze_impactanalyze_changesget_index_statssearch_contentsearch_knowledgeread_knowledge_documentAdvanced primitives are hidden from tools/list by default. Set PITLANE_MCP_TOOL_TIER=all to expose the full surface.
The pitlane binary exposes the same code intelligence outside MCP.
Examples:
pitlane index /your/project
pitlane investigate /your/project "How does ignore handling work?"
pitlane analyze-changes /your/project --base-ref main --include-working-tree
pitlane search /your/project authenticate --kind method
pitlane symbol /your/project src/auth.rs::Auth::login[method]
pitlane usages /your/project src/auth.rs::Auth::login[method]
All commands output JSON to stdout. Logs go to stderr and are controlled with RUST_LOG.
pitlane-mcp supports Rust, Python, JavaScript, TypeScript, Svelte, C, C++, Go, Java, C#, Bash, Ruby, Swift, Objective-C, PHP, Zig, Lua, Kotlin, and Solidity.
For the full extension matrix and symbol-kind coverage, see Languages and Symbol Kinds.
pitlane-mcp is feature-complete and maintained as an OSS navigation tool.Licensed under either of MIT License or Apache License, Version 2.0, at your option.
Token-efficient navigation substrate for AI coding agents. Index code once and retrieve only the symbols you need.
工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Pitlane 透過 stdio 在本機執行。
上次檢查時間:2026-11-05。 Token-efficient navigation substrate for AI coding agents. Index code once and retrieve only the symbols you need.