# pitlane-mcp

> Token-efficient navigation substrate for AI coding agents. Index code once and retrieve only the symbols you need.

- **URL**: https://www.freshcrate.ai/projects/pitlane-mcp
- **Author**: eresende
- **Category**: MCP Servers
- **Latest version**: `v0.10.3` (2026-05-18)
- **License**: Apache-2.0
- **Source**: https://github.com/eresende/pitlane-mcp
- **Language**: Rust
- **GitHub**: 60 stars, 9 forks
- **Registry**: github (`eresende/pitlane-mcp`)
- **Tags**: `ai`, `code-navigation`, `coding-agent`, `developer-tools`, `llm`, `mcp`, `model-context-protocol`, `rust`

## Description

Token-efficient navigation substrate for AI coding agents. Index code once and retrieve only the symbols you need.

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v0.10.3` | 2026-05-18 | High | ## pitlane-mcp v0.10.3 - bounded session state, cached investigations, and agent ergonomics  This patch release focuses on making the v0.10.x agent-navigation surface steadier under repeated use. It bounds per-session memory growth, improves repeated `investigate` calls, tightens regex search guardrails, and makes agent-facing output easier to read.  ### Bounded project session state  pitlane now caps the number of project session records retained in memory and prunes older entries with an |
| `v0.10.2` | 2026-05-08 | High | ## pitlane-mcp v0.10.2 - source symlink hardening and dependency updates  This patch release tightens source-file path handling after the v0.10.x navigation and indexing work. It makes the source discovery policy explicit: pitlane rejects symlink-backed source files instead of following them during indexing, searching, or read-time access.  Thanks to @tamlut-modnys for the symlink-hardening contribution in #33.  ### Reject symlinked source files  pitlane now rejects source files whose path is ba |
| `v0.10.1` | 2026-04-25 | High | ## pitlane-mcp v0.10.1 — cache correctness, watcher search freshness, and safer repository walks  This patch release tightens the indexing pipeline after the v0.10.0 navigation overhaul. It focuses on making repeated indexing reliably use the cache, ensuring watch-mode updates correctly refresh BM25 search, and improving behavior on large or messy repositories with generated/vendor directories and unreadable paths.  ### Correct index cache freshness  `index_project` now records enough sour |
| `v0.10.0` | 2026-04-20 | High | ## pitlane-mcp v0.10.0 — intent-aware navigation, single-call investigation, and session-aware reads  This release is the biggest architectural step since v0.9.0. It introduces a new high-level tool tier built around intent rather than implementation detail, a single-call investigation tool that replaces multi-step discovery sequences, and a session layer that tracks what the agent has already seen so it can steer toward expansion instead of redundant rereads.  ### `investigate` — answer a c |
| `v0.9.1` | 2026-04-15 | High | # pitlane-mcp v0.9.1  This patch release focuses on making startup behavior more predictable, preventing semantic-search hangs, and improving the project documentation and demo material.  ## Highlights  - `ensure_project_ready` is now non-blocking with respect to embeddings. - Semantic `search_symbols` queries now have a timeout to avoid hanging on slow embedding backends. - The CLI now reports total embeddings stored separately from the number newly written during the current run. - Th |
| `v0.9.0` | 2026-04-10 | Medium | ## pitlane-mcp v0.9.0 - smarter startup, better discovery, Solidity support, and safer workspace boundaries  ### Smarter startup and discovery  This release makes pitlane much easier to boot up and navigate from an agent workflow.  A new `ensure_project_ready` tool provides a preferred one-call startup path by combining indexing with conditional embedding waits, so agents can get to a usable project state with less orchestration.  Discovery also got a major upgrade. `search_content` adds |
| `v0.8.0` | 2026-04-08 | Medium | ## pitlane-mcp v0.8.0 - graph-aware navigation, better indexing, and stronger Svelte support  ### Graph-aware navigation  `find_callees` and `find_callers` are now available in the MCP server, giving pitlane shallow graph-aware navigation without changing its focused scope.  Use `find_callees` to inspect a symbol’s direct outgoing references, and `find_callers` to see direct incoming references for quick impact checks. This is meant to stay lightweight and fast: enough graph awareness to i |
| `v0.7.0` | 2026-04-06 | Medium | ## pitlane-mcp v0.7.0 - BM25 camelCase tokenizer and better indexing  ### BM25 camelCase tokenizer  The BM25 index now uses a custom tokenizer that splits at camelCase, PascalCase, and digit boundaries before indexing. `LowerInstruction` is indexed as `["lower", "instruction"]`, `parseHTTPResponse` as `["parse", "http", "response"]`, `base64Encode` as `["base", "64", "encode"]`. Queries like `"lower instruction"` or `"http response"` now match symbols they previously missed entirely. This fi |
| `v0.6.0` | 2026-04-04 | Medium | ## What's New  ### New Tools  - **`get_lines`** — fetches a file slice by line range for blocks that aren't named symbols. Useful when an agent needs context around a specific line without reading the whole file. - **`get_index_stats`** — returns symbol counts by language and kind. Lets agents orient in a new codebase without burning tokens on `get_project_outline`.  ### Enhancements  - **`get_symbol` signature-only default** — when fetching a class, struct, interface, or trait symbol |
| `v0.5.0` | 2026-04-03 | Medium | ## What's New    ### New Languages    - **Bash** — tree-sitter indexing for `.sh` and `.bash` files. Extracts functions.   - **C#** — tree-sitter indexing for `.cs` files. Extracts classes, structs, interfaces, enums, methods, constructors,   delegates, and type aliases. Class and struct bodies are trimmed to the header line when methods are present.   - **Ruby** — tree-sitter indexing for `.rb` files. Extracts classes, modules, methods, and singleton methods. Class and   module bodies a |

## Citation

- HTML: https://www.freshcrate.ai/projects/pitlane-mcp
- Markdown: https://www.freshcrate.ai/projects/pitlane-mcp.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/pitlane-mcp/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._