Cursor + GitHub + Vercel: The Complete AI Dev Pipeline

The dream of AI-assisted development is Cursor creating a branch, writing the fix, opening a PR, and triggering a staging deploy — while you review the output. That workflow exists today. The part most people skip is configuring the permission layer so each step requires exactly the right amount of human oversight.

This guide walks through the full Cursor + GitHub + Vercel pipeline via Stow, with a permission configuration that makes the useful parts automatic and the risky parts require your approval.

How the Architecture Works

Every tool call Cursor makes — whether to GitHub or Vercel — routes through Stow's policy layer first. You configure each service independently, and Stow enforces those decisions before anything reaches the API.

Why Each Service Needs Its Own Permission Set

GitHub Risk

Merging to main, deleting branches, triggering Actions without review

Vercel Risk

Triggering production builds, modifying env vars, managing domains

Cursor (Agent) Risk

Chaining operations across services faster than you can track

Step 1: Connect All Three Services in Stow

• In Connected Services, add GitHub with a Personal Access Token scoped to repos and workflows
• Add Vercel with a Vercel API Token
• Both appear as separate services in your dashboard with independent permission controls

Step 2: Configure GitHub Permissions

Step 3: Configure Vercel Permissions

Step 4: Connect Cursor to Stow

• In your Stow dashboard, go to AI Agents → + Add Agent
• Name it "Cursor" and click New Client Secret — copy the Agent ID and Secret
• In Cursor, open Settings (Cmd/Ctrl+Shift+J) and go to Tools & MCP → + New MCP Server
• Add the Stow SSE URL with your credentials:

{
  "mcpServers": {
    "stow": {
      "type": "sse",
      "url": "https://mcp.stowprotect.com/sse?agent_id=YOUR_AGENT_ID&agent_secret=YOUR_SECRET"
    }
  }
}

Step 5: The Complete Pipeline Workflow

Prompts That Drive the Pipeline

What Gets Logged