Skip to content
Codemap
Esc
navigateopen⌘Jpreview
On this page

Comparison

Where Codemap fits among code-index and agent-context tools — by product category, not a feature scorecard.

An agent needs “who imports this?” or “which tests are affected?” — some tools answer with SQL, some with Cypher, some with fixed MCP verbs, some with a quality-gate binary. They are not the same product. This page sorts peers by what they’re trying to be, then scores the axes Codemap targets.

Static linters, LSP servers, and whole-repo embeddings sit in other slots. See When to skip.

Four kinds of “code intelligence”

Category Exemplars Compare on
Queryable structural index Codemap SQL + recipes, apply, CI formats, freshness
Polyglot knowledge graph codebase-memory-mcp Language breadth, embeddings, Cypher, cross-repo
Quality-gate analyzer fallow Dead-code depth, CI verdicts, framework rules
Fixed-tool / visual explorer codegraph · codedb · code-graph (FalkorDB) MCP context packs · snapshot search · 3D/GraphRAG

What Codemap targets (and what it delegates)

Codemap owns structural facts you can SELECT, not semantic ranking or editor UX. Embeddings, hover/rename, and verdict autofix stay with peers in those categories. Opt-in FTS5 (--with-fts) JOINs body text with structure — it is not an embedding index.

Cells below were checked against peer source (tool registries / store DDL), not README claims alone. Last verified: 2026-07-20.

Capability Codemap CBM fallow codegraph codedb code-graph
Category Queryable index Polyglot knowledge graph Quality-gate analyzer Fixed-tool explorer Fixed-tool explorer Visual / GraphRAG
Persistent embedded store ✅ SQLite ✅ SQLite (+ .zst art.) 🟡 analysis cache (not SQL) ✅ SQLite 🟡 custom snapshot (not SQL) 🟡 FalkorDB or embedded lite
Ad-hoc query language ✅ SQL + recipes 🟡 Cypher + fixed tools ❌ fixed tools / flags ❌ fixed tools (explore default) ❌ fixed MCP + query pipeline ❌ fixed MCP (Cypher internal)
Apply / edit from index ✅ recipe · rows · diff 🟡 removal-oriented fix 🟡 fallback edit
CI formats (SARIF / badge / …) ✅ broad (incl. PR/review)
Agent init / install wiring 🟡 template init
Index freshness / watch ✅ git-poll 🟡 re-analyze watch ❌ no continuous watcher
Language breadth 🟡 JS/TS/CSS + text ✅ 163 grammars 🟡 JS/TS/web (oxc) ✅ WASM tree-sitter polyglot ✅ Zig-native polyglot 🟡 few langs (JS yes, TS no)
Embeddings / semantic search ❌ FTS5 opt-in only ✅ on-device vectors ❌ FTS/structural ❌ trigram/word indexes ❌ (GraphRAG chat ≠ index emb.)
Visual graph UI 🟡 mermaid ✅ 3D ✅ 3D
Local, no phone-home ✅ opt-in off 🟡 default-on telemetry 🟡 default-on (opt-out flag)

Legend: ✅ full · 🟡 partial · ❌ none. CBM = codebase-memory-mcp. codegraph defines eight MCP tools; default ListTools is explore only unless the allowlist is widened. Engine / agent / CI axes only — not a full quality-gate scorecard for fallow.

Differentiator: SQL + named recipes as the query API, a general apply engine over explicit diff rows, and CI format breadth — on an embedded SQLite index with CLI / MCP / HTTP parity. Peers that win languages, embeddings, or context-packing usually do not ship that combination.

By design, not missing

  • No embeddings in the box — structural over semantic (product tenets). Reach for a knowledge-graph peer when you need cosine search across many languages.
  • No verdict engine as a primitive — SARIF / annotations are output modes of recipes; codemap apply runs rows you provide.
  • No mandatory graph server — FalkorDB-style hosting is a different install story; Codemap stays local-first.
  • JS/TS/CSS depth over polyglot breadth (v1) — honest ceiling; see When to skip.

When to pick which

You want… Reach for
SQL queries, apply, and CI gates on a JS/TS/CSS tree Codemap (@stainless-code/codemap)
163 languages, on-device embeddings, Cypher over a SQLite graph codebase-memory-mcp
Deep JS/TS quality gates (dead code, dupes, security candidates) in CI fallow
Single-call / explore-style MCP context packing codegraph (default tool: explore)
Zig-native MCP context engine (trigram + snapshot; edit as fallback) codedb
Interactive 3D graph + HTTP GraphRAG over a multi-lang repo code-graph
Grep semantics, LSP rename, or knip-style autofix Those tools — When to skip

Same name, different product

Other GitHub projects are also called “codemap.” They are not this package:

  • AZidan/codemap — Python symbol → line-range indexer (JSON under .codemap/); CLI only; cheaper full-file reads, not SQL.
  • JordanCoin/codemap — Go project-brain / tree / handoff CLI (+ MCP); often installed as Homebrew/Scoop codemap. Session priming, not a queryable structural SQLite schema.

Install this one as @stainless-code/codemap. Positioning without peer names: Why Codemap.

Cells re-checked against peer source on 2026-07-20. Re-verify before treating them as everlasting.

Last updated on July 20, 2026

Was this page helpful?