AI analysis grounded in the code graph — computed facts, not vibes · 2026-07-05T09:38:00Z
codebase-memory-mcp is an MCP (Model Context Protocol) server, written in C, that indexes a codebase into a persistent knowledge graph so AI coding agents can query code structure rather than reading files individually. It parses source via tree-sitter across a claimed 158 languages (see cbm_language_for_extension in src/discover/language.c) and layers "Hybrid LSP" semantic type resolution on top for 9–11 named languages, with per-language cross-reference passes such as cbm_batch_go_lsp_cross, cbm_batch_php_lsp_cross, and cbm_batch_c_lsp_cross. The output is a graph of functions, classes, call chains, and routes exposed through 14 MCP tools, plus an optional 3D graph UI (GraphTab.tsx, NodeDetailPanel.tsx). It targets developers running agentic coding workflows who want faster, cheaper structural context.
The star count is high (26,577) but this repository gained 0 stars this period, so there is no growth to explain here — the momentum, whatever produced the existing base, has stalled during the observed window. The README leans heavily on quantified claims (Linux kernel indexed in 3 minutes, sub-1ms queries, 5,604 passing tests, an arXiv preprint reporting 83% answer quality and 10× fewer tokens across 31 repos), which plausibly explains the accumulated star base, but the provided facts give no evidence of a current growth catalyst.
What changed recently, how it's actually built (from the code graph), and whether you should care. Free account — no card, no spam.