The codebase-memory-mcp repository is a high-performance code intelligence server built in C that indexes entire codebases into a persistent knowledge graph. It is designed to serve as an MCP (Model Context Protocol) server for AI coding agents, enabling rapid structural code queries and analysis. The project ships as a single static binary with zero dependencies, available for macOS, Linux, and Windows, making deployment straightforward without requiring Docker, runtime dependencies, or API keys.
The core capability is extreme indexing speed. The repository can index an average codebase in milliseconds and processes the Linux kernel with 28 million lines of code across 75,000 files in approximately 3 minutes. Queries against the indexed knowledge graph execute in under 1 millisecond. This performance is achieved through a RAM-first pipeline using LZ4 compression, in-memory SQLite, and fused Aho-Corasick pattern matching, with memory released after indexing completes.
Language support spans 158 programming languages through vendored tree-sitter grammars compiled directly into the binary. Beyond generic AST parsing, the system implements Hybrid LSP semantic type resolution for ten languages including Python, TypeScript, JavaScript, PHP, C#, Go, C, C++, Java, Kotlin, and Rust. This produces a knowledge graph containing functions, classes, call chains, HTTP routes, and cross-service links with type-aware accuracy comparable to major language servers.
The repository provides 14 MCP tools covering search, trace, architecture analysis, impact analysis, Cypher-like queries, dead code detection, cross-service HTTP linking, and ADR (Architecture Decision Records) management. Semantic search uses bundled Nomic nomic-embed-code embeddings with 11-signal combined scoring including TF-IDF, RRI, API signatures, AST profiles, data flow analysis, and graph diffusion. BM25 full-text search operates via SQLite FTS5 with a camelCase-aware tokenizer.
The system detects and links cross-service communication patterns including HTTP routes, gRPC, GraphQL, tRPC, and pub-sub channels across Socket.IO, EventEmitter, and generic patterns in eight languages. Infrastructure-as-code indexing handles Dockerfiles, Kubernetes manifests, and Kustomize overlays as graph nodes with cross-references. A built-in 3D interactive graph visualization UI runs at localhost:9749 for exploring the knowledge graph.
The most active labels are bug (241 items), parsing/quality (98 items), and priority/normal (95 items). The repository overlaps contributors with genmcp/gen-mcp, knative/eventing, and golang/go.
The installation process is automated through a single command that auto-detects and configures 11 coding agents including Claude Code, Cursor, VS Code, Aider, KiloCode, and others. The project includes a preprint research paper on arXiv (2603.27277) documenting design and benchmarks showing 83% answer quality, 10 times fewer tokens, and 2.1 times fewer tool calls compared to file-by-file exploration across 31 real-world repositories.
