grafio-mcp 0.3.0

Structural code knowledge graph for AI coding assistants — impact analysis, cross-service flows, and symbol search via MCP

Grafio

Structural code knowledge graph for AI coding assistants.

Grafio indexes your codebase into a symbol-level graph and exposes it via MCP (Model Context Protocol) so AI assistants like Claude can understand cross-service flows, find callers, and analyze impact — without reading every file.

Install

pip install grafio              # the daemon
npm install -g grafio-mcp       # the MCP bridge for Claude Code

Quick start

# Start the daemon (auto-starts when Claude Code opens if grafio is on PATH)
grafio serve

# Index a repo
grafio index --repo-id my-project --root /path/to/my-project

Configure Claude Code — create ~/.claude/.mcp.json:

{
  "mcpServers": {
    "grafio": {
      "command": "grafio-mcp",
      "env": {
        "GRAFIO_URL": "http://127.0.0.1:9742"
      }
    }
  }
}

Restart Claude Code. Run /mcp to verify.

What it does

• code_graph_context — understand a method + everything it calls across files
• code_graph_impact — find what breaks if you change a symbol
• code_graph_search — find symbols by name or REST endpoint path
• code_graph_update — incrementally reindex changed files

Claude automatically decides when to use Grafio vs Grep+Read based on the question.

Supported languages

Java, TypeScript, JavaScript, Python, Go, Rust

License

MIT