codebase-memory-mcp (DeusData)
https://github.com/DeusData/codebase-memory-mcp
High-performance MCP server building persistent code knowledge graphs via AST. 155 languages via vendored tree-sitter; SQLite + WAL; single static binary; zero runtime deps. Knowledge graph: Project/Package/File/Module/Class/Function/Method/Interface/Enum/Type/Route/Resource nodes; CALLS/IMPORTS/DEFINES/IMPLEMENTS/INHERITS/HTTP_CALLS/ASYNC_CALLS/EMITS/LISTENS_ON/DATA_FLOWS/SIMILAR_TO/SEMANTICALLY_RELATED edges. 14 MCP tools incl. Cypher-like query.
At a glance
- Type
- Tree-sitter knowledge-graph MCP for code
- Tier
- T1
- Section
- Claude Code memory mechanisms
- Created
- 2026 (active 2026 release cadence; latest v0.6.1 2026-05-04)
- Latest release
- v0.6.1 / 2026-05-04
- License
- MIT
- Pricing
- free / open source (MIT); no paid tier
- Funding
- searched not found
Taxonomy
- storage
- graph
- retrieval
- structural-query
- persistence
- project-bound
- update
- replacement
- unit
- fact
- governance
- user-controllable
- conflict
- deterministic-overwrite
When to use
Optimised for: structural code knowledge graphs (token-efficient; 99.2% reduction vs grep) + sub-ms queries + 155-language tree-sitter coverage
Anti-fit: not for non-code memory; not for natural-language conversation memory; SEMANTICALLY_RELATED edges require optional embedding pipeline
Pros & cons
Pros
Single-binary zero-deps deployment; 155-language tree-sitter coverage; structural queries replace grep at 120x fewer tokens; SLSA-3 + Sigstore + VirusTotal; team-shared graph artifact format; preprint with reproduction-friendly evaluation.
Cons
Preprint not peer-reviewed; founders/company opaque (DeusData has no disclosed principals); no managed/cloud option; SEMANTICALLY_RELATED edges depend on optional embedding pipeline.
Claims & capabilities
Linux kernel (28M LOC, 75K files) indexed in 3 min; sub-ms queries; 99.2% token reduction (3,400 vs 412,000); 120x fewer tokens vs grep; preprint reports 83% answer quality + 2.1x fewer tool calls across 31 repos
Technical surface
- API surface
- MCP (stdio) — 14 tools incl. index_repository / search_graph / trace_call_path / query_graph (Cypher-like) / get_architecture / detect_changes / manage_adr / search_code / get_code_snippet / ingest_traces / list_projects / delete_project / index_status / get_graph_schema
- Backend storage
- SQLite (WAL mode) at ~/.cache/codebase-memory-mcp/
- Deployment
- Local — single static binary (macOS arm64/amd64; Linux arm64/amd64; Windows amd64)
- Embedding model
- not applicable — structural graph (no embedding model required); SEMANTICALLY_RELATED edges optional
- Multi-tenancy
- not applicable — single-user local
- MCP
- native (first-party MCP server)
- A2A
- not supported
- OpenTelemetry
- searched not found
Compare codebase-memory-mcp (DeusData) with…
Similar systems
Other claude code memory mechanisms in the catalog, ranked by inbound references.
- Official MCP Memory server T1
Knowledge-graph-based persistent memory MCP server — the canonical reference implementation. npx -y @modelcontextprotocol/server-memory ; configurable via MEMORY_FILE_PATH .
- Anthropic Auto Dream T1
Background sub-agent in Claude Code that periodically scans ~/.claude/projects/<project>/memory/ + recent JSONL transcripts via grep, prunes stale notes, deduplicates, resolves contradictions, converts relative dates to absolute timestamps, and rewrites MEMORY.md (kept under 200 lines). Triggers automatically when 24+ hours elapsed AND 5+ sessions completed; manually via /dream.
- Claude Code auto-memory T1
Persistent file-based memory at ~/.claude/projects/<project>/memory/ with a MEMORY.md index plus typed memory files (user / feedback / project / reference). Loaded into context across sessions.
- claude-brain T2
Claude Code plugin packaging of Memvid (see Dedicated memory layers). Wraps the Memvid .mv2 single-file store as a slash-command-installable plugin so per-project memory becomes a git-committable file inside the repo.
- Claude-CursorMemoryMCP T2
FastAPI server. RBAC + API-key auth, encryption, backup/recovery, embedding pipeline with deduplication. Only server in this list with role-based access control over pgvector.
- claude-mem T1
Hooks into Claude Code's session lifecycle; compresses tool outputs into ~500-token observations via Claude's agent SDK. SQLite + FTS5 with optional Chroma vector. "Endless Mode" compresses in-flight to extend session length. Captures what the agent did , not just what the user typed.
Related systems
References (2)
- Anthropic Claude (foundation models) depends on at runtime — adjacent-infrastructure cell: self-contained binary + SQLite WAL; MCP client (Claude Code/Codex CLI/etc.); no LLM lock-in
- Model Context Protocol (MCP spec) depends on at runtime — adjacent-infrastructure cell: self-contained binary + SQLite WAL; MCP client (Claude Code/Codex CLI/etc.); no LLM lock-in