mcpdirectory.dev
9,129 servers · updated 23 Sept

Graqle MCP server

community Active last commit 47 d 
{
  "mcpServers": {
    "graqle": {
      "command": "uvx",
      "args": [
        "graqle"
      ]
    }
  }
}
file .mcp.json (in your project root)install uvx graqleversion 0.27.0

What it does

GraQle is an MCP server that indexes codebases as knowledge graphs to enable AI agents to reason about architecture and make auditable decisions. It serves developers building AI-assisted code tools and organizations needing cryptographically verifiable records of AI decisions in production.

  • Query a codebase as a typed graph to find security risks, architectural violations, or code patterns across files.
  • Capture and anchor production AI decisions (loan approvals, hiring recommendations, medical triage) to a public transparency log for regulatory compliance.
  • Teach custom rules into a codebase graph so that future audits and code generation automatically enforce architectural constraints.

Tools this server exposes

ToolDoes
scanIndex a codebase into a typed knowledge graph with functions, classes, modules, imports, and calls.
runExecute multi-agent reasoning across the graph to audit it (e.g., find authentication bypass risks) and return confidence scores with evidence trails.
learnPersist custom rules into the graph so future audits and code generation enforce them automatically.
attestCapture an AI decision with metadata (domain, model_id, inputs, output) and produce a tamper-evident record.
pseudonymize_refRedact personally identifiable information from decision inputs before attestation.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.