🧠 Connect Your AI

Step 3: Connect Your AI Subscription

Mission Control is bring-your-own-AI. Pick a provider, connect it once, and your squad runs on it. We don't mark up tokens.

Pick your provider

You pick the provider on the setup page. This guide explains how to get whichever credential the page asks for.

---

ChatGPT (recommended)

Before you start

1. Get a flat-rate ChatGPT plan ($100 or $200/mo) - the tiers sized for squad workloads
2. Enable "Device code authorization for Codex" in your ChatGPT Security settings

How it works

1. On the setup page, click ChatGPT
2. Submit the form. Your workspace provisions first.
3. After provisioning, the success screen shows a device code and an authorization link
4. Open the link, paste the code, click Approve
5. Done. No tokens to copy or store on your end.

ChatGPT issues OAuth credentials directly to your workspace. No API key handling.

---

Claude (Extra Usage required)

⚠️ Heads up: As of April 4, 2026, Claude subscriptions no longer cover third-party usage like Mission Control. You need Extra Usage enabled, billed at API prices. For flat-rate, use ChatGPT.

Get your setup token

1. Subscribe to Claude Max at claude.ai/settings/billing and enable Extra Usage
2. Open a terminal in FULL SCREEN and run:

npx @anthropic-ai/claude-code setup-token

3. Browser opens. Click Approve to authorize.
4. Token appears in your terminal, starting with sk-ant-oat01-...
5. Copy the entire token and paste it on the setup page

Token format

sk-ant-oat01-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

⚠️ Copying tips:

• If the terminal window is narrow, the token may wrap onto two lines. Copy the entire token.
• Paste into a text editor first, remove any extra spaces or line breaks, then copy that into the setup page.
• The token should be one continuous string with no line breaks.

Alternative: Anthropic API key

You can also paste an Anthropic API key directly (starts with sk-ant-api03-...) for pay-per-token billing without the Claude Pro/Max subscription path. Pick API key instead of Setup token on the setup page.

Platform notes

# Mac (Terminal):
npx @anthropic-ai/claude-code setup-token

# If you get permission errors (EACCES), don't run npx with sudo -
# fix the npm cache ownership once, then re-run the command:
sudo chown -R $(whoami) ~/.npm

# Windows (PowerShell):
npx @anthropic-ai/claude-code setup-token

# If npx isn't recognized:
npm exec @anthropic-ai/claude-code setup-token

---

MiniMax

How it works

1. Subscribe to the MiniMax Coding Plan
2. On the setup page, click MiniMax and submit the form
3. Your workspace provisions first
4. After provisioning, connect your MiniMax account in AI Tokens with one click

---

Z.AI

Get your API key

1. Subscribe to the GLM Coding Plan
2. Create an API key at Z.AI API Keys
3. On the setup page, click Z.AI and paste the API key

---

Claude Code (your Claude plan as a full agent)

Claude Code runs your agents through the Claude Code CLI right on your workspace computer, on the Claude subscription you already pay for. Plan limits only: it never bills extra usage.

1. Run claude setup-token in a terminal (or npx @anthropic-ai/claude-code setup-token)
2. Approve in the browser and copy the token (starts with sk-ant-oat01-...)
3. In AI Tokens, open the Claude Code card and paste the token

Connecting verifies with a real test turn from your workspace, so a bad token fails at the modal, never on an agent's first task. One account per workspace; reconnecting replaces it.

---

Grok

1. In AI Tokens, click Sign in with Grok on the Grok card
2. Approve the code on grok.com
3. Done. The connection keeps itself fresh; if it ever lapses, the dashboard tells you.

Works with SuperGrok and X Premium subscriptions. No API key.

---

Gemini

1. In AI Tokens, click Sign in with Google on the Gemini card
2. Approve, then paste the redirect URL back if the browser cannot reach your workspace directly

The free tier works; AI Pro or Ultra lifts the limits. For chat-speed replies pick Gemini 3 Flash: the free tier throttles Gemini 3 Pro heavily.

---

GitHub Copilot

1. In AI Tokens, click Sign in with GitHub on the Copilot card
2. Enter the device code on github.com

One Copilot plan fronts Claude, GPT, and Gemini models. Your plan's actual entitlements appear in the model picker after connecting.

---

Kimi

1. Subscribe to a Kimi coding plan (K3 or K2.7)
2. Create an API key in the Kimi console
3. Paste it on the Kimi card in AI Tokens

---

Qwen Cloud

1. Get the API key from your Qwen plan's console (token plan or coding plan)
2. Paste it on the Qwen Cloud card in AI Tokens

Connecting verifies your key with a real call. The plan includes Qwen 3.8 Max, the 3.7 family, and cross-vendor picks (DeepSeek V4 Pro, GLM-5.2) through the same subscription.

---

OpenRouter

1. Create a key at openrouter.ai
2. Paste it on the OpenRouter card in AI Tokens

Every major vendor's models become one dropdown away, billed to your OpenRouter balance. Handy as a cross-vendor fallback chain too.

---

Troubleshooting