codebase-memory-mcp
codebase-memory-mcp is a code intelligence engine for AI coding agents. It parses your repository with tree-sitter, enhanced by language-server type resolution for major languages, and stores the result as a persistent knowledge graph. Agents then answer structural questions such as who calls this function, what breaks if I change this, or where the entry points are, with one graph query instead of dozens of grep and file-read cycles.
It suits developers working in medium to very large codebases, where agents waste most of their tokens just finding their way around. The project reports indexing an average repository in milliseconds and the Linux kernel in about three minutes, and in its own benchmark five structural queries used around 3,400 tokens versus roughly 412,000 with file-by-file search. It exposes 17 MCP tools, indexes infrastructure files such as Dockerfiles and Kubernetes manifests, and includes a 3D graph viewer. The design is described in an arXiv preprint.
Getting started: run the one-line installer on macOS or Linux, or the PowerShell installer on Windows. It downloads a single static binary (no Docker, runtime or API key) and configures the coding agents it detects. Restart your agent and ask it to index the project.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-22
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 13
- Latest release
- v0.12.1 (2026-10-16)
- Open issues
- 640
- Recent issues with at least one comment
- 2/12
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #1654 0.10.4 `index_repository` OOM-crashes on large repo — v0.9.0 indexed the same tree in ~13 min*24 comments · 2026-08-15
- #221 "install" command does not work for opencode in windows 1123 comments · 2026-04-07
- #1145 Windows-Indexing worker crashed on a file17 comments · 2026-07-17
- #45 Very high CPU usage16 comments · 2026-03-11
- #1582 Claude.ai Desktop App still cannot start mcp15 comments · 2026-08-13
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- Your agent burns tokens searching a medium or large codebase: one graph query replaces many grep and read steps, according to the README.
- You want everything local: a single static binary with no runtime dependencies or API keys.
- You work in many languages and with several coding agents, which the installer configures automatically.
- You do not want an installer to edit your agents' config files and add hooks; the README offers --skip-config to install only the binary.
- You need project decisions and conventions rather than code structure; a scaffold such as mex is closer.
- You are cautious about unsigned binaries: some antivirus engines have flagged the release, which the project documents as a false positive.
How it differs
vs gortex★ 1.9kBoth build a code knowledge graph for coding agents. Codebase Memory is MIT with a single binary and a stated focus on token savings. Gortex lists more MCP tools (49) and multi-repo workspaces, but GitHub does not detect its licence, so confirm the terms first.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!