codebase-memory-mcp vs graphify
Code intelligence MCP in pure C: tree-sitter knowledge graph over 158 languages, average repo indexed in milliseconds, sub-ms queries, 10x fewer tokens. Single static binary, zero deps. — versus — /graphify skill for Claude Code, Codex, Cursor and 20+ assistants: maps code, docs, PDFs and media into a local knowledge graph — tree-sitter AST, EXTRACTED/INFERRED edges, no vector store.
Same job — a tree-sitter code knowledge graph agents query to save tokens. codebase-memory-mcp is a zero-dep C binary over MCP with sub-ms queries; graphify is Python, code-plus-docs, and labels communities and god nodes for orientation.
| codebase-memory-mcp | graphify | |
|---|---|---|
| Stars | 45k | 122k |
| Forks | 3.7k | 12k |
| Language | C | Python |
| License | MIT | Apache-2.0 |
| Last activity | 4 days ago | yesterday |
| Topics | code-intel, local | code-intel, knowledge-graphs, skills |
| Curated connections | 10 | 5 |
codebase-memory-mcp — the curator's take
The performance ceiling of the code-context-server category: pure C, single static binary, the Linux kernel indexed in 3 minutes, structural queries under a millisecond — with a peer-reviewed preprint (83% answer quality, 10x fewer tokens across 31 repos) instead of vibes. Hybrid LSP adds real type resolution for the 12 languages that matter most, and 43 client surfaces means it plugs into whatever agent you run. NOT semantic search — it answers structural questions (call chains, routes, blast radius), not 'where's the code that does X'; pair it with an embedding tool for that. And note its own disclosure: it writes to your agent config files by design — audit posture is unusually good (SLSA 3, OpenSSF, per-release VirusTotal), use it.
graphify — the curator's take
Reach for graphify when your agent burns tokens re-reading a repo to answer 'how does X connect to Y': one run writes graph.json plus a GRAPH_REPORT.md, then `graphify query`, `path` and `explain` answer from the graph, and strict mode in Claude Code redirects the session's first raw file read to it. Code extraction is free and deterministic (tree-sitter, ~40 languages, Leiden communities, zero LLM credits); docs, PDFs and video cost model tokens for the semantic pass, so a docs-heavy corpus is not free. Every edge is tagged EXTRACTED or INFERRED — trust the first, verify the second. Skip it if you want an always-on MCP code server with blast-radius tooling (code-review-graph, codebase-memory-mcp) or symbol-level edits (serena): graphify is on-demand, and the background-sync version is the paid graphify.com platform. Disclosure: StackMap's own discovery layer runs on graphify.