Codemesh
Codemesh builds a persistent knowledge graph of your codebase and exposes it through MCP tools or direct CLI calls. Instead of grepping and reading dozens of files, agents call codemesh_answer or codemesh_explore to get structured context about symbols, dependencies, and call paths. The graph supports semantic search with embeddings, PageRank scoring for relevance, and write-back operations so agents can enrich the graph with what they learn. Benchmarks show 71% cost reduction and 72% faster execution versus grep-and-read baselines across six real repositories. It runs as an MCP server for structured tool workflows or as a subprocess for zero-overhead Bash integration. Best for large codebases where discovery overhead dominates agent token budgets.
- tools
- 10
- transport
- STDIO
- env var
- 1
- last updated
- Apr 20, 2026
Tools
10 toolsPublic tool metadata for what this MCP can expose to an agent.
codemesh_mapBounded, one-level navigation through directories, files, symbols, and graph links
Bounded, one-level navigation through directories, files, symbols, and graph links
No parameter schema in public metadata yet.
codemesh_sourceExact source for a selected symbol ID
Exact source for a selected symbol ID
No parameter schema in public metadata yet.
codemesh_inspectExact source for a symbol plus direct call evidence, within a byte and symbol budget
Exact source for a symbol plus direct call evidence, within a byte and symbol budget
No parameter schema in public metadata yet.
codemesh_hydrateVerify saved symbol references and reopen exact source after context loss
Verify saved symbol references and reopen exact source after context loss
No parameter schema in public metadata yet.
codemesh_answerOptional ranked context assembly with source snippets
Optional ranked context assembly with source snippets
No parameter schema in public metadata yet.
codemesh_exploreSearch, context (multi-target), impact analysis
Search, context (multi-target), impact analysis
No parameter schema in public metadata yet.
codemesh_traceFollow call chains with source code
Follow call chains with source code
No parameter schema in public metadata yet.
codemesh_enrichWrite back what you learned for future sessions
Write back what you learned for future sessions
No parameter schema in public metadata yet.
codemesh_workflowRecord multi-file workflow paths
Record multi-file workflow paths
No parameter schema in public metadata yet.
codemesh_statusGraph health check
Graph health check
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageConfiguration
CODEMESH_PROJECT_ROOTAbsolute path to the local codebase to index
Readme
Codemesh
Intelligent code knowledge graph for AI coding agents
71% cheaper, 72% faster, 82% fewer tool calls vs baseline Grep+Read
on 6 real-world repos (Sonnet 4.6) — from a single codemesh index.
Benchmarks · Quick Start · Integrations · Write-Back · How It Works · API Reference · Full Results
The Problem
AI coding agents waste 40-80% of their tokens on discovery — grepping through files, reading irrelevant code, and rebuilding context they've already seen in previous sessions.
Details
- Repository
- github.com/pyalwin/codemesh
- Category
- Documents & Knowledge
- Package
@pyalwin/codemesh- Transport
- STDIO
- Registry
- MCP registry, active
Related Documents & Knowledge MCP servers
View all
Rarasrarasai/rarasDisease Twin — Brazil's rare-disease knowledge graph (10,468) over MCP. Source-grounded, PT-BR.0stars
Adaptive Memory Graphraskolnikovdd/adaptive-memory-graphPersistent memory for Claude via weighted, interconnected knowledge nodes.0stars
KnowDriverecourse-software-inc/knowdrive-mcpPrivate, multimodal knowledge store for agents: hybrid search with cited, provenance-backed results.0stars
bookstore4agentsscott2121/bookstore4agentsMarketplace where AI agents buy and sell books — curated domain knowledge that improves tasks.0stars
Hive Zk Attestationsrotzin/hive-mcp-zk-attestationZero-knowledge attestation for agent capability and trust score claims0stars