Skip to content

Connect Remote MCP ​

Access your CQ knowledge base from any MCP-compatible AI — ChatGPT, Claude Desktop, Cursor — without installing CQ locally.


What You'll Do ​

Connect ChatGPT (or any MCP client) to your CQ AI Workspace via mcp.pilab.kr. After setup, you can search knowledge, save snapshots, and check project status from anywhere.


Step 1: Log In ​

Visit mcp.pilab.kr and log in with GitHub. This creates your OAuth token.

sh
cq auth login    # If you haven't already

Step 2: Add MCP Server to Your AI Tool ​

Claude Desktop / Claude Code ​

Add to your MCP config:

json
{
  "cq-brain": {
    "url": "https://mcp.pilab.kr/mcp",
    "type": "streamable-http"
  }
}

ChatGPT ​

In ChatGPT settings → MCP Servers → Add:

URL: https://mcp.pilab.kr/mcp

ChatGPT will walk through OAuth — log in with GitHub when prompted.

Cursor ​

In .cursor/mcp.json:

json
{
  "mcpServers": {
    "cq-brain": {
      "url": "https://mcp.pilab.kr/mcp",
      "type": "streamable-http"
    }
  }
}

Step 3: Use It ​

Once connected, your AI has access to these tools:

ToolWhat it does
cq_knowledge_recordSave a conversation snapshot to your knowledge base
cq_recallSearch your knowledge base
cq_statusCheck project status

Example: Search knowledge from ChatGPT ​

"Search my CQ knowledge for WebSocket retry patterns"

ChatGPT calls cq_recall → returns your past decisions and code patterns.

Example: Save an idea from Claude Desktop ​

"Save this conversation as a snapshot — we figured out the caching strategy"

Claude calls cq_knowledge_record → stored in Supabase, searchable from any tool.


How It Works ​

Your AI tool → mcp.pilab.kr/mcp
                    │
                    ▼
             Cloudflare Worker (OAuth proxy)
                    │
                    ▼
             Supabase Edge Function (MCP server)
                    │
                    ▼
             Your knowledge base (Supabase DB)

No CQ binary needed on the remote machine. The AI Workspace lives in the cloud.


Next Steps ​