codebase-memory-mcp codebase-memory-mcp
codebase-memory-mcp is a high-performance code intelligence MCP server that indexes a repository into a persistent knowledge graph using tree-sitter across 162 languages, with hybrid LSP type resolution for major languages. Coding agents query the graph through 17 MCP tools — search, call tracing, architecture overviews, impact analysis, dead-code detection, Cypher queries — replacing dozens of grep/read cycles with a few structural queries.
The project surpassed 44,000 GitHub stars and trended on GitHub in September 2026; its design and benchmarks were published as an arXiv preprint (2603.27277) reporting evaluation across 31 real-world repositories, and recent releases added a shared session coordination daemon and Kubernetes/Kustomize indexing.
As of
- Indexes an average repo in milliseconds and the 28M-LOC Linux kernel in about three minutes with a RAM-first pipeline
- 162 vendored tree-sitter grammars compiled into one zero-dependency binary
- Hybrid LSP semantic type resolution for Python, TypeScript/JavaScript, Go, Rust, Java, Kotlin, C/C++, C#, PHP and Perl
- 17 MCP tools including impact analysis, dead code detection, cross-service HTTP linking, ADR management and Cypher queries
- `install` auto-configures Claude Code, Codex, Cursor, OpenCode and 40+ other client surfaces with hooks and skills
- Built-in 3D graph visualisation UI served from the binary at localhost:9749
- Infrastructure-as-code indexing for Dockerfiles, Kubernetes manifests and Kustomize overlays
Developers running Claude Code, Codex, Cursor or other MCP-capable agents on large or polyglot codebases who want structural code understanding with far lower token usage.
Very small projects where plain file search is already cheap, or environments that forbid running third-party native binaries that modify agent configuration files.
Open source under the MIT license; distributed as a free single static binary for macOS, Linux and Windows (also on npm, PyPI and Go). No API keys, accounts or paid tiers are required.
A mature, well-benchmarked way to give coding agents a real map of a codebase; the single-binary install and broad client coverage make it one of the easiest code-intelligence MCP servers to adopt, with the usual caveat of trusting a tool that writes to your agent config.
Is codebase-memory-mcp free?
Yes - codebase-memory-mcp is Open Source. Open source under the MIT license; distributed as a free single static binary for macOS, Linux and Windows (also on npm, PyPI and Go). No API keys, accounts or paid tiers are required.
Is codebase-memory-mcp open source?
Yes - codebase-memory-mcp is open source. Open source under the MIT license; distributed as a free single static binary for macOS, Linux and Windows (also on npm, PyPI and Go). No API keys, accounts or paid tiers are required.
Who is codebase-memory-mcp best for?
Developers running Claude Code, Codex, Cursor or other MCP-capable agents on large or polyglot codebases who want structural code understanding with far lower token usage.
Who is codebase-memory-mcp not ideal for?
Very small projects where plain file search is already cheap, or environments that forbid running third-party native binaries that modify agent configuration files.
What are the best codebase-memory-mcp alternatives?
The closest codebase-memory-mcp alternatives on ai.dosa.dev are Greptile, Fern, remio - all listed under AI Codebase Knowledge & Generation.
Who makes codebase-memory-mcp?
codebase-memory-mcp is developed by DeusData. It is listed in the AI Codebase Knowledge & Generation category on ai.dosa.dev.
Favorite this tool to revisit it later, or Zap it to contribute to the public vote count.
Content on this page is AI-generated. Please verify details with the vendor's website for accuracy.