Skip to content
A-Team

Developers

Connect your AI assistant. Describe what to build.

A-Team MCP gives capable coding assistants a governed interface to learn the platform, create solutions, validate, deploy, test, inspect, and iterate.

Public MCP · use any assistant

Build A-Team solutions with the AI you already use.

Connect Claude Code, ChatGPT, Cursor—any MCP assistant—to A-Team. It reads the spec, designs your solution, deploys, and tests while you describe the outcome. No lock-in: every solution is Git-backed and yours to keep.

Setup · 2 steps

  1. 1

    Create a workspace

    Spin up a tenant on the A-Team platform.

    Create workspace →
  2. 2

    Copy your workspace key

    Paste it into your assistant when it asks—that's the only thing it needs to connect.

    adas_<workspace>_<key>

Then just tell your assistant: “Connect to A-Team MCP.”

A-Team MCP

ChatGPT connects to A-Team as tools and builds a full solution for you.

See it in Claude Code

Add the MCP once, then just describe the outcome. Claude Code connects to A-Team and builds it—reading the spec, deploying, and testing—right in your terminal.

claude code — ~/support-bot

Watch Claude Code connect to A-Team and build a solution from one sentence.

● Try it free — no account

Connect your assistant and start exploring

Point any MCP client at A-Team with no key. It bootstraps, learns the platform, and drafts a working solution with you. When you're ready to validate and deploy, it prompts you to create a free workspace—grab your tenant key from Tokens & Keys in the app and paste it when asked.

Remote endpoint: https://mcp.ateam-ai.comnpm: @ateam-ai/mcp

ChatGPT

Developer Mode connector
Settings → Connectors → Developer Mode
Paste:  https://mcp.ateam-ai.com

Claude Code

one command
claude mcp add ateam -- npx -y @ateam-ai/mcp

Claude Desktop · Cursor · VS Code

mcp.json
{
  "mcpServers": {
    "ateam": {
      "command": "npx",
      "args": ["-y", "@ateam-ai/mcp"]
    }
  }
}

Works with any MCP-compatible client (Gemini, Copilot, Windsurf, …) over Streamable HTTP or stdio.

Quick start

Five steps to your first system.

  1. 1Get an A-Team API key.
  2. 2Add the MCP endpoint to your assistant.
  3. 3Ask it to bootstrap and inspect examples.
  4. 4Describe the desired system.
  5. 5Review, deploy, and test.
a-team · mcp

$ connect a-team mcp

→ bootstrap · learn the spec

→ inspect examples

→ describe the system

→ review architecture

→ deploy · test · improve

✓ solution deployed

Mental model

Four concepts to hold.

Solution

The complete application topology.

Skill

An operational AI role.

Connector

A controlled MCP integration.

Governance

Permissions, contracts, policies, and handoffs.

Lifecycle

Learn → build → observe → version.

Learn
Build
Validate
Deploy
Test
Observe
Patch
Version

Get an API key

Request access to A-Team MCP and start building with your assistant. Key issuance and client-specific configuration are documented once verified.