@thinkingpatterns/mcp-server

MCP server that exposes ThinkingPatterns security findings management as tools for AI coding agents.

Works with any coding agent that supports the Model Context Protocol — Claude Code, Cursor, Cline, Windsurf, and others.

Quick start

npm install -g @thinkingpatterns/mcp-server

Get an API key from ThinkingPatterns Settings, then follow the setup guide for your agent:

| Agent | Guide | |-------|-------| | Claude Code | docs/claude-code.md | | Cursor | docs/cursor.md | | Cline | docs/cline.md |

What it provides

14 MCP tools for managing security findings, scans, metrics, and analysis agents. See the full tool reference for parameters and return types.

Findings — list_findings, get_finding, triage_finding, bulk_triage, add_comment

Scanning — list_runs, ingest_sarif

Projects & Metrics — list_projects, get_metrics, get_top_risk_modules

Agents — list_agents, get_agent, create_agent, run_agent

Project context

Drop a .thinkingpatterns.json in your repo root to auto-associate the project:

{
  "project_id": "<your-project-uuid>"
}

Your coding agent can read this file to avoid calling list_projects at the start of each session.

Architecture

The package ships as a stdio MCP server — your coding agent spawns it as a child process and communicates over stdin/stdout using the MCP protocol. It calls the ThinkingPatterns API via @thinkingpatterns/core.

Coding Agent  ←— stdio (MCP protocol) —→  thinkingpatterns-mcp  ——→  ThinkingPatterns API

Docs

• Tool reference — every tool, parameter, and return type
• Claude Code setup
• Cursor setup
• Cline setup