AST knowledge graph MCP server for Claude Code — semantic search, call graph traversal, HTTP route tracing, impact analysis. Auto-indexes 10 languages via Tree-sitter.
How to install code-graph-mcp?
sdsrss/code-graph-mcp
$git clone https://github.com/sdsrss/code-graph-mcpInstalls into the current project.
›Prefer a prompt? Paste this to your agent
Install code-graph-mcp by running `git clone https://github.com/sdsrss/code-graph-mcp`, then use it for the current task and follow its documentation at https://github.com/sdsrss/code-graph-mcp.
README.md
| 1 | # code-graph-mcp |
| 2 | |
| 3 | A high-performance code knowledge graph server implementing the [Model Context Protocol (MCP)](https://modelcontextprotocol.io/). Indexes codebases into a structured AST knowledge graph with semantic search, call graph traversal, and HTTP route tracing — designed to give AI coding assistants deep, structured understanding of your code. |
| 4 | |
| 5 | ## Features |
| 6 | |
| 7 | - **Multi-language parsing** — Tree-sitter AST extraction across tiers of depth: |
| 8 | - **Full** (calls + imports + inheritance + test markers): TypeScript/TSX, JavaScript, Go, Python, Rust, Java. HTTP route extraction additionally covers TypeScript/TSX + JavaScript (Express/Connect), Go (`net/http`), and Python (Flask/FastAPI) only — Rust and Java web frameworks are not yet route-extracted |
| 9 | - **Smoke-tested** (calls + imports + inheritance): C#, Kotlin, Ruby, PHP, Swift, Dart |
| 10 | - **Limited** (functions + calls + `#include` imports + gtest test markers + C++ base-class inheritance; `Class::method` scope qualification deferred): C, C++ |
| 11 | - **Scripting**: Bash (functions + commands + `source`/`.` imports), Markdown (headings) |
| 12 | - **File-FTS only** (no AST symbol extraction): HTML, CSS, JSON |
| 13 | - **Semantic code search** — Hybrid BM25 full-text + vector semantic search with Reciprocal Rank Fusion (RRF), powered by sqlite-vec |
| 14 | - **Call graph traversal** — Recursive CTE queries to trace callers/callees with cycle detection |
| 15 | - **HTTP route tracing** — Map route paths to backend handler functions (Express, Flask/FastAPI, Go `net/http`) |
| 16 | - **Dead code detection** — Find unreferenced symbols with smart Orphan/Exported-Unused classification |
| 17 | - **Impact analysis** — Determine the blast radius of code changes by tracing all dependents |
| 18 | - **Incremental indexing** — Merkle tree change detection with file system watcher for real-time updates. Smart event filtering skips metadata-only changes (chmod, xattr) |
| 19 | - **Context compression** — Token-aware snippet extraction for LLM context windows (L0→full code, L1→summaries, L2→file groups, L3→directory overview). Compact JSON output saves 15-20% tokens |
| 20 | - **Embedding model** — Optional local embedding via Candle (feature-gated `embed-model`). Context reordered to prioritize structural relations over code for better embedding quality |
| 21 | - **Self-healing** — Automatic SQLite corruption recovery with rebuild. Startup repair for incomplete indexing (Phase 3 failures) |
| 22 | - **MCP protocol** — JSON-RPC 2.0 over stdio, plug-and-play with Claude Code, Cursor, Windsurf, and other MCP clients |
| 23 | - **Claude Code Plugin** — First-class plugin with slash commands (`/understand`, `/trace`, `/impact`), agents, skills, auto-indexing hooks, StatusLine integration, and self-updating |
| 24 | |
| 25 | ## Why code-graph-mcp? |
| 26 | |
| 27 | Unlike naive full-text search or simple AST dumps, code-graph-mcp builds a **structured knowledge graph** that understands the relationships between symbols across your entire codebase. |
| 28 | |
| 29 | ### Incremental by Design |
| 30 | |
| 31 | BLAKE3 Merkle tree tracks every file's content hash. On re-index, only changed files are re-parsed — unchanged directory subtrees are skipped entirely via mtime cache. When a function signature changes, **dirty propagation** automatically regenerates context for all downstream callers across files. |
| 32 | |
| 33 | ### Hybrid Search, Not Just Grep |
| 34 | |
| 35 | Combines BM25 full-text ranking (FTS5) with vector semantic similarity (sqlite-vec) via **Reciprocal Rank Fusion (RRF)** with raw score blending — so searching "handle user login" finds the right function even if it's named `authenticate_session`. Results are auto-compressed to fit LLM context windows. |
| 36 | |
| 37 | ### Scope-Aware Relation Extraction |
| 38 | |
| 39 | The parser doesn't just find function calls — it tracks them within their proper scope context. Extracts calls, imports, inheritance, interface implementations, exports, and HTTP route bindings. Same-file targets are preferred over cross-file matches to minimize false-positive edges. |
| 40 | |
| 41 | ### HTTP Request Flow Tracing |
| 42 | |
| 43 | Unique to code-graph-mcp: trace from `GET /api/users` → route handler → service layer → database call in a single query. Supports Express, Flask/FastAPI, and Go HTTP frameworks. |
| 44 | |
| 45 | ### Zero External Dependencies at Runtime |
| 46 | |
| 47 | Single binary, embedded SQLite, bundled sqlite-vec extension, optional local embedding model via Candle — no database server, no cloud API, no Docker required. Runs entirely on your machine. |
| 48 | |
| 49 | ### Built for AI Assistants |
| 50 | |
| 51 | Every design decision — from token-aware compression to node_id-based snippet expansion — is optimized for LLM context windows. Works out of the box with Claude Code, Cursor, Windsurf, and any MCP-compatible client. |
| 52 | |
| 53 | ## Performance |
| 54 | |
| 55 | | Metric | Value | |
| 56 | |--------|-------| |
| 57 | | Indexing speed | **300+ files/second** (single-threaded, release build) | |
| 58 | | Incremental re-index | **<250ms** no-change detection via BLAKE3 Merkle tree | |
| 59 | | FTS search P50 / P99 | **<300us / <1ms** | |
| 60 | | Database overhead | **~3.5MB** per 800 nodes | |
| 61 | | Token savings | **5-20x fewer tokens** per code understanding task vs grep+read | |
| 62 | |
| 63 | Run `code-graph-mcp benchmark` on |
Related
6 picksType
- code-review-graphLocal-first knowledge graph for token-efficient code review through MCP and CLIJul 2026
- repowiseCodebase intelligence for AI coding agents — graph, git history, docs, decisions, code health.Jul 2026
- mcp-server-atlassian-bitbucketNode.js/TypeScript MCP server for Atlassian Bitbucket. Enables AI systems (LLMs) to interact with workspaces, repositories, and pull requests via tools (list, get, comment, search). Connects AI directly to version control workflows through the standard MCP interface.Mar 2026
- gemini-mcp-toolMCP server for Gemini CLI integrationJul 2026
- socraticodeEnterprise-grade (40m+ LOC) codebase intelligence, zero-setup, local & private Plugin/Skill/Extension or MCP: hybrid semantic search, polyglot dependency graphs, symbol-level impact analysis & call-flow, interactive HTML viewer, cross-project & branch-aware search, DB/API/infra…Jul 2026
- github-mcpProduction-ready Model Context Protocol server providing Claude with complete GitHub integration - comprehensive tools for repository management, branch operations, file handling, issue tracking, and collaboration features with OAuth 2.0 authentication and MongoDB persistence.Jun 2026
