Open sourceC · ≈48k stars · MIT
codebase-memory-mcp
Local code-intelligence MCP server: a persistent graph of code, docs and infrastructure for AI agents.
codebase-memory-mcp is an MIT-licensed MCP server, written in C, that indexes code, documentation, configuration and infrastructure into a persistent local knowledge graph. Its 17 MCP tools combine symbol, text and semantic search with callers, dependencies and file outlines, and Hybrid LSP adds type resolution. It traces relationships through Docker, Kubernetes, Terraform and CI config, and supports change impact and test selection. The installer configures dozens of agent clients and adds skills and context hooks. An optional visual companion, CBM Lens, is in beta. Analysis runs locally.
Pros & cons
Each point cites its sources and says what kind of evidence backs it. A second reviewer tried to refute every claim before it was kept.
Pros
- DocsA single native C binary with vendored parsers for 162 languages and bundled code embeddings; indexing needs no Docker, API key or Ollama.[1]
- User reportsUsers say its index cuts full-file reads, and one says it dramatically improved output quality.[1][2][3]
- Docs17 MCP tools covering graph, text and semantic search, call paths and change impact, with an installer for 45 client integrations.[1]
- MetadataMIT licensed and very active (v0.12.1, Oct 2026).[1]
Cons
- Independent testIn one user's test on the Pi harness, it raised token use to 85k input tokens. The same task took 80k with no tool and 67k with a PROJECT.md pointer in AGENTS.md, at the same output quality.[1]
- User reportsA rival tool's maintainer benchmarked retrieval over 7,213 queries across 14 repos. Codebase-memory scored 0.502 MRR, behind LemonCrow (0.676) and cocoindex-code (0.557).[1]
- DocsCall paths come from static analysis and show possible relationships, not what actually ran, unless you ingest traces separately.[1]
- Metadata475 open issues on a repo created in February 2026.[1]
- Best for
- Large or multi-language codebases where agents waste turns grepping and reading whole files.
- Watch out for
- Benchmark token use and answer quality with and without it on your own repo; on small projects it can add overhead.
Assessed Oct 29, 2026 · adversarially reviewedSuggest a correction
People
Founder
Creator of codebase-memory-mcp (DeusData)
MCP & tools · Context engineering & memory
In research (4)
- Stratum: Repository Memory for Coding Agents Beyond the Context WindowVerdict: Incremental
Oct 26, 2026 · A code-graph MCP server, close to Stratum's structural memory tier exposed via MCP tools.
- RepoTrace: Improving Requirement Tracing in Multi-Round Agent InteractionsVerdict: Incremental
Oct 26, 2026 · A code-graph context tool that tackles the same repository-understanding problem.
- How Do Small Coding Agents Use Code Graphs and Embeddings?Verdict: Important
Oct 25, 2026 · A code-graph MCP server of exactly the kind whose premise the paper tests.
- Code Comprehensor: Hierarchical Documentation and Cross-Modal Retrieval for Large RepositoriesVerdict: Incremental
Oct 21, 2026 · Open-source MCP code-graph server that does the same job (repository structure for agents).
Neighbours on the workbench
TypeScript · ≈393k stars · MIT
openclaw/openclaw
The viral open-source personal AI assistant that really does things.
Maintained by OpenClaw Foundation
MDX · ≈307k stars · MIT
obra/superpowers
An agentic skills framework and software methodology that works.
Maintained by Jesse Vincent (obra)
Shell · ≈306k stars · MIT
mattpocock/skills
Small, composable agent skills for real engineering work.
Maintained by Matt Pocock (AI Hero)
Connections
Relations from the Harness Hub knowledge graph, as charted in the State of Harness reports.
Maintained by · 1
- DeusDataOrganisation
Built on · 1
- Model Context ProtocolProject
Similar entries · 2
- SerenaProject
- SourcegraphStartup
Sources
Verified Oct 29, 2026