Control Plane MCP Server — Overview

2026-05-14 · source · source ↗ · raw capture

"Connect AI assistants and tools to Control Plane using the Model Context Protocol (MCP)"

Server Details

Quick Start

  1. Create a Service Account — generate an API key; add to "superusers" or "viewers" groups
  2. Configure Your Client — tool-specific setup for Antigravity, Claude Code, Codex, Cursor, Gemini CLI, VS Code
  3. Start Building — interact with Control Plane through natural language

Service Account Permissions

Two built-in groups: - viewers — view on all resources; read-only exploration - superusers — manage on all resources; full automation

Custom permissions: create groups with specific members, define policies with targeted permissions.

Recommended setups: - Quick exploration → viewers - Full dev access → superusers - Production automation → custom group with specific policies - Team access → group per team with policies scoped to GVCs

Initialization Flow

  1. Read the Guide (cpln+virtual://guide) — tool taxonomy and best practices
  2. Read Server Info (cpln+virtual://about) — server version, capabilities
  3. Set Context (set_context) — establish default org and GVC
  4. Begin Operations — use 80+ tools

AI Plugin

"Auto-configures this MCP server and adds 23 skills, 8 agents, 8 commands, and 8 guardrail rules."

Best Practices

  1. Start with Context — always set org and GVC at conversation start
  2. Be Specific — provide specific values for memory, CPU, replicas
  3. Confirm Destructive Actions — configure assistant to confirm before deletes/updates
  4. Use Tags — add tags for organization and filtering

Recommended prompt structure: Context → Action → Target → Details

Compatible Clients

Any MCP remote-server-compatible client. Explicit guides for: Antigravity, Claude Code CLI, Codex, Cursor IDE, Gemini CLI, VS Code.