Your codebase has a shape.
Now your AI can see it.

GraphMind indexes your code into a knowledge graph and exposes it to your AI assistant via MCP. Ask about dead code, cross-project dependencies, or why that service exists — and get answers grounded in your actual architecture. No noise, no raw file dumps — up to 5,700x fewer tokens than unstructured search. Your AI shouldn't have to rediscover your codebase every session.

Works with Claude Cursor Windsurf Cline Zed Continue · any MCP client
CLI / Developer
graphmind — index
View on GitHub ↗
Desktop App · Mac & Windows
GraphMind desktop app — ProjectsGraphMind desktop app — IntegrationsGraphMind desktop app — Settings
5,700x
less noise for your AI
99.9%
token reduction vs raw search
25
MCP tools, zero setup
~10M
tokens saved per session

Three steps to a smarter AI assistant

Works the same whether you use the desktop app or the CLI.

Desktop App Mac & Windows · no terminal required
01

Download & open

Install the GraphMind app for Mac or Windows. Open it, point it at your project folder. No terminal needed.

02

Connect your AI assistant

The app configures your AI tools automatically. One click — the MCP server starts and your assistant sees your graph immediately.

03

Ask anything

"What calls this function?" "Show dead code." "Why does this service exist?" Get answers grounded in your actual code.

or via CLI
CLI / Developer Homebrew · Cargo · full control
01

Index your codebase

graphmind index .

GraphMind parses every file with tree-sitter, builds a symbol graph in DuckDB, and detects cross-project dependencies automatically.

02

Connect to your AI assistant

graphmind setup

One command configures everything: MCP server in Claude Desktop, Claude Code, Cursor, Windsurf — and injects usage rules into your global context. 25 tools available immediately.

03

Ask anything

"What calls this function?" "Show dead code." "Why does this service exist?" Get answers grounded in your actual code.

What your AI assistant can do with your graph

25 MCP tools. All grounded in your actual code.

gm_dead_code

Dead code detection

Find functions with no callers across all your projects. Safe to delete with confidence.

gm_fn_impact

Blast radius

Know what breaks before you touch it. Full transitive caller chain before you touch anything.

gm_diff_impact

Diff impact analysis

Run on any git diff before merging. See exactly which symbols are affected and who calls them.

gm_memory_add

Architectural memory

Store decisions, patterns, conventions. Your AI assistant recalls them at every session, forever.

gm_similar

Duplication detection

Find structurally similar symbols across your codebase. Spot copy-paste patterns before they become debt.

gm_cross_links

Cross-project links

See exactly how your services depend on each other. Detected automatically at index time.

gm_cycles

Circular dependency detection

Find dependency cycles before they cause build failures or runtime surprises.

gm_export

Graph export

Export any file or symbol neighborhood as Mermaid or DOT. Drop it in your PR or architecture doc.

gm_search

Hybrid search

Find code by meaning, not just by name. FTS + semantic + graph, all combined.

Benchmark · 31K symbol codebase

Your AI gets the answer, not the noise

Raw code search dumps millions of tokens of irrelevant lines into your AI's context. GraphMind returns ranked symbols with structural context — up to 5,700x less noise. On a typical session with 5–10 searches, that's ~10M tokens saved.

Query grep tokens GraphMind tokens Ratio
"payment processing" ~1,468,000 ~257 5,700x
"compliance check" ~1,007,000 ~274 3,670x
"wallet creation" ~1,237,000 ~285 4,340x

* 1 token ≈ 4 characters · benchmark on a 31K symbol codebase

Beyond token count

grep GraphMind
Understands synonyms "money transfer" → payment_service
Ranks by relevance RRF (3 sources)
Structural context callers/callees via graph
Filters noise AST symbols only
~10M
tokens saved per session
On 5–10 searches vs grep. Nearly an entire context window.
5,700x
peak reduction ratio
1,468,000 grep tokens → 257 GraphMind tokens.

Simple, transparent pricing

Start free. Upgrade when you need more.

MonthlyAnnualSave 20%

Free

0€
  • Local graph, unlimited projects
  • 25 MCP tools
  • SQLite memory store
  • Local embeddings (minilm)
  • MIT — open source

Embeddings

9€/mo
  • Everything in Free +
  • Remote semantic search
  • Voyage AI embeddings
  • Better cross-session results
Most popular

Pro

19€/mo
  • Everything in Embeddings +
  • Remote API + MCP server
  • No local installation needed
  • Works from any machine

Team

15€/mo/seat
  • Everything in Pro +
  • Shared team graph
  • Shared architectural memories
  • gm_team_who_knows — find who documented a symbol
  • Auto-sync after indexing
  • Min. 2 seats

Trusted by engineers who take their codebase seriously

"GraphMind changed how I onboard new devs. I just point my AI assistant at the graph and let it explain the architecture."

— CTO, fintech startup

"Dead code detection alone saved us two weeks of refactoring. We deleted 3,000 lines with confidence."

— Staff Engineer

"The memory feature is what I didn't know I needed. My AI assistant now knows our conventions without me repeating them."

— Senior Developer

Start in 30 seconds

Free forever. No account needed. Open source.

Download for Mac / Windows brew install graphmind

Click the command to copy

View on GitHub