Install Hatchik — connect your AI to your stack

Grab your Hatchik API key

Easiest: it’s already in your sandbox-ready email — the one we send the moment you sign up. The full MCP snippet is there with the key pre-filled. Skip to step 4 if so.

Or create a fresh one at hatchik.com/account → API keys. The plaintext is shown once, copy it then.

Copy this config

Replace YOUR_HATCHIK_API_KEY with the key from step 1. This same JSON works for every AI tool we support — only the file you paste it into changes (step 3).

mcp config — same for all tools

{
  "mcpServers": {
    "hatchik": {
      "command": "npx",
      "args": ["-y", "hatchik-mcp"],
      "env": {
        "HATCHIK_API_KEY": "YOUR_HATCHIK_API_KEY",
        "HATCHIK_API_URL": "https://api.hatchik.com"
      }
    }
  }
}

Add it to your AI tool

Find your tool below and follow the steps. Most tools have a UI panel for MCP servers — paste the snippet there. The terminal tools (Claude Code, Codex) keep their config in a file.

Tool	What to do
Cursor	Settings → MCP → Add server → paste the snippet. Save.
Windsurf	Click the Cascade icon in the side panel → tools icon at the top → Configure → paste the snippet. Save.
Cline	In VS Code: click the Cline icon → MCP settings panel → paste the snippet.
Antigravity	Settings → MCP Servers → Add → paste the snippet.
Claude Code	Open or create ~/.claude/mcp.json and paste the snippet as the whole file (or merge if the file already exists).
Claude Desktop	Open or create ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows) and paste the snippet.
Codex	Open or create ~/.codex/mcp.json and paste the snippet.
Continue	Open or create ~/.continue/config.json and paste the snippet.
Aider	Add it under aider.mcp.servers in your aider config.
Copilot (agent mode)	In VS Code at your repo root: open or create .vscode/mcp.json and paste the snippet.

How do I open or create the file paths above? (Finder & File Explorer tips)

The ~/ at the start of a path means “your home folder”. So ~/.cursor/mcp.json is inside your home folder, in a hidden folder called .cursor, and the file is called mcp.json.

On a Mac (Finder):

Open Finder.
Press Cmd+Shift+. (period) — this reveals hidden folders.
Press Cmd+Shift+G, paste the path, hit Go. If the folder doesn’t exist, Finder will offer to create it.
If the file isn’t there, create it with TextEdit (save as plain text, not rich text).

On Windows (File Explorer):

Open File Explorer.
View menu → tick “Hidden items”.
Click the address bar, paste the path (replace ~ with %USERPROFILE%), hit Enter.
If the file isn’t there, right-click → New → Text Document, rename to mcp.json (turn on file extensions if hidden).

Using something else MCP-capable that isn’t listed? Try it — if it speaks the Model Context Protocol, the same JSON works. Email us where you put it and we’ll add a row.

Using ChatGPT, GPT, Gemini, or Grok in a browser?

Quick clarification: GPT, Gemini, and Grok are AI models, not tools. You can’t directly “install” the MCP snippet for them — there’s no config file. Instead:

If you mean the desktop tool that uses these models: ChatGPT → the closest fit is the Codex CLI (already in the table above); Gemini → Antigravity (Google’s IDE, in the table). Wire those up the normal way.
If you mean the browser chat (chat.openai.com, gemini.google.com, grok.com): MCP doesn’t apply. Open the chat, attach AI_CONTEXT.md from your repo, paste this as your first message: “Read AI_CONTEXT.md and let’s start.” You’ll copy edited files back into your repo yourself — or use the deploy URL/token from AI_CONTEXT.md to push without a terminal.
If you mean the AI inside your app (chatbots, summaries, etc.): different question entirely. Put your provider key in your app’s .env — works with Anthropic, OpenAI, Google, xAI (Grok), Mistral, OpenRouter.

Full per-tool walk-through at docs.hatchik.com/ai-tool-setup.

Restart your AI tool, then ask

Quit and reopen so it picks up the new MCP config. Then in chat:

// You:
Are you connected to Hatchik? Show me my recent deploys.

Your AI should answer with the project name, the sandbox URL, the region, and your last few deploys. If it says it has no Hatchik tools — the config didn’t load. Re-check the file path, double-check the API key for typos, and that you fully quit the app before reopening.

What this does, in plain English: you’re telling your AI tool there’s a service called “hatchik” it’s allowed to talk to. When you ask it to ship a feature, check why the site is slow, or roll back yesterday’s change, it can do that directly — the AI calls our MCP server, our server checks your API key, and the action runs against your Hatchik project. Anything that costs money or could break things (payments, deletes, rollbacks) still bounces through your browser for a click-to-confirm.

Plug your AI into your Hatchik.

Grab your Hatchik API key

Copy this config

Add it to your AI tool

Using ChatGPT, GPT, Gemini, or Grok in a browser?

Restart your AI tool, then ask

Stuck on a step?