Atlas / MCP servers / optave / Codegraph

CodegraphBLOCK

Name: Codegraph
Author: optave

mcp/optave/codegraph-5

MCP server by optave · ai agents · architecture · ci cd · Audited 2026-10-07

Code intelligence CLI — function-level dependency graph across 34 languages, 34-tool MCP server for AI agents, complexity metrics, architecture boundary enforcement, CI quality gates, git diff impact with co-change analysis, hybrid semantic search. Fully local, zero API keys required.

Verdict

BLOCK

Grade

Trust score

38 /100

Exposed tools

49 42r · 6w · 1d

Transport

stdio

License

Apache-2.0

Stars

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

codegraph

Give your AI the map before it starts exploring.

= 22.12.0" />

The Problem · What It Does · Quick Start · Commands · Languages · AI Integration · How It Works · Practices · Roadmap

The Problem

AI agents face an impossible trade-off. They either spend thousands of tokens reading files to understand a codebase's structure — blowing up their context window until quality degrades — or they assume how things work, and the assumptions are often wrong. Either way, things break. The larger the codebase, the worse i

Read from source at commit f3a3aaadf990OBSERVED · 2026-10-07

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.

claude-code

claude mcp add lib --env CODEGRAPH_LLM_API_KEY=${CODEGRAPH_LLM_API_KEY} -- npx -y @myorg/[email protected]

claude-desktop

{
  "mcpServers": {
    "lib": {
      "command": "npx",
      "args": [
        "-y",
        "@myorg/[email protected]"
      ],
      "env": {
        "CODEGRAPH_LLM_API_KEY": "${CODEGRAPH_LLM_API_KEY}"
      }
    }
  }
}

Exposed tools (49)

42 read · 6 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.

Tool	Risk	Description
`ast_query`	read	Search stored AST nodes (calls, literals, new, throw, await) by pattern. Requires a prior build.
`audit`	read	Composite report combining explain, fn-impact, and health metrics for a file or function. Returns structure, blast radius, complexity, and threshold breaches in one call.
`batch_query`	write	Run a query command against multiple targets in one call. Returns all results in a single JSON payload — ideal for multi-agent dispatch.
`branch_compare`	write	Compare code structure between two git refs (branches, tags, commits). Shows added/removed/changed symbols and transitive caller impact using temporary git worktrees.
`brief`	read	Token-efficient file summary: symbols with roles and transitive caller counts, importer counts, and file risk tier (high/medium/low). Designed for context injection.
`cfg`	read	Show intraprocedural control flow graph for a function.
`check`	write	CI gate: run manifesto rules (no args), diff predicates (with ref/staged), or both (with rules flag). Returns pass/fail verdicts.
`co_changes`	write	Find files that historically change together based on git commit history. Requires prior
`code_owners`	read	Show CODEOWNERS mapping for files and functions. Shows ownership coverage, per-owner breakdown, and cross-owner boundary edges.
`communities`	read	Detect natural module boundaries using Leiden community detection
`complexity`	read	Show per-function complexity metrics (cognitive, cyclomatic, nesting, Halstead, Maintainability Index). Sorted by most complex first.
`config`	read	Show or manage codegraph configuration (project + user-level global config)
`context`	read	Full context for a function: source code, dependencies with summaries, callers, signature, and related tests — everything needed to understand or modify a function in one call
`cycles`	read	Detect circular dependencies in the codebase
`dataflow`	read	Show data flow edges or data-dependent blast radius.
`diff_impact`	read	Analyze git diff to find which functions changed and their transitive callers
`execution_flow`	read	Trace execution flow forward from an entry point through callees to leaves, or list all entry points with list=true
`export`	read	Export dependency graph as DOT, Mermaid, JSON, GraphML, GraphSON, or Neo4j CSV
`export_graph`	read	Export the dependency graph in DOT, Mermaid, JSON, GraphML, GraphSON, or Neo4j CSV format
`file_deps`	read	Show what a file imports and what imports it
`file_exports`	read	Show exported symbols of a file with per-symbol consumers — who calls each export and from where
`find_cycles`	read	Detect circular dependencies in the codebase
`fn_impact`	read	Show function-level blast radius: all functions transitively affected by changes to a function
`impact_analysis`	read	Show files affected by changes to a given file (transitive)
`implementations`	read	List all concrete types (classes, structs, records) that implement a given interface or trait
`info`	read	Show codegraph engine info and diagnostics
`interfaces`	read	List all interfaces and traits that a given class, struct, or record implements
`list`	read	List all registered repositories
`list_functions`	read	List functions, methods, classes, structs, enums, traits, records, and modules in the codebase, optionally filtered by file or name pattern
`list_repos`	read	List all repositories registered in the codegraph registry
`map`	read	High-level module overview with most-connected nodes
`mcp`	write	Start MCP (Model Context Protocol) server for AI assistant integration
`models`	read	List available embedding models
`module_map`	read	Get high-level overview of most-connected files
`node_roles`	read	Show node role classification (entry, core, utility, adapter, dead [dead-leaf, dead-entry, dead-ffi, dead-unresolved], leaf) based on connectivity patterns
`path`	read	Find shortest path between two symbols (or files with file_mode) in the dependency graph
`plot`	read	Generate an interactive HTML dependency graph viewer
`prune`	destructive	Remove stale registry entries (missing directories or idle beyond TTL)
`query`	read	Query the call graph: find callers/callees with transitive chain, or find shortest path between two symbols
`registry`	read	Manage the multi-repo project registry
`roles`	read	Show node role classification: entry, core, utility, adapter, dead (dead-leaf, dead-entry, dead-ffi, dead-unresolved), leaf
`semantic_search`	read	Search code symbols by meaning using embeddings and/or keyword matching (requires prior
`sequence`	read	Generate a Mermaid sequence diagram from call graph edges. Participants are files, messages are function calls between them.
`snapshot`	write	Save and restore graph database snapshots
`stats`	read	Show graph health overview: nodes, edges, languages, cycles, hotspots, embeddings
`structure`	read	Show project structure with directory hierarchy, cohesion scores, and per-file metrics. Per-file details are capped at 25 files by default; use full=true to show all.
`symbol_children`	read	List sub-declaration children of a symbol: parameters, properties, constants. Answers
`triage`	read	Ranked audit queue by composite risk score (connectivity + complexity + churn + role)
`where`	read	Find where a symbol is defined and used, or list symbols/imports/exports for a file. Minimal, fast lookup.

Trust audit

BLOCKgrade F · trust 38/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.

Layer	What it checks	Result
L0	Provenance & inventory	WARN
L1	Static analysis of the code	FAIL
L2	Instruction surface (what it tells the agent)	FAIL
L3	Class-specific surface	WARN
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (16 observation(s))
Network: declared (3 observation(s))
Shell: declared (8 observation(s))
Dependencies: not all pinned
Secrets in source: none-found

Findings (25)

CRITICALPrompt injection · prompt.transfer_instruction · CWE-94, CWE-1427

.claude/skills/sweep/SKILL.md:88

All subagents in this sweep — and any other concurrent session — share one authenticated `gh` identity's rate limit (5000 REST requests/hour). Polling too tightly, or re-fetching every endpoint on eve

Why it matters. an instruction to move sensitive data to an outside destination

Fix. remove; a skill never needs the user's secrets off the machine