Getting started
You need a Tiny Command workspace and an MCP-compatible host (Claude Desktop or Cursor). The whole setup is three steps.
In the platform, open MCP keys and run the setup wizard. Pick a permission preset (the Standard preset lets your AI build, save, and operate everything), name the key, and generate it.
Your secret key is shown once, so copy it immediately. It looks like tcm_live_aB3xK9mPqRv7….
Paste this block into your host's MCP configuration, replacing the placeholder with your key.
Claude Desktop (claude_desktop_config.json):
{
"mcpServers": {
"tinycommand": {
"url": "https://mcp.tinycommand.com/v1",
"headers": {
"Authorization": "Bearer YOUR_KEY_HERE"
}
}
}
}
Cursor (~/.cursor/mcp.json) uses the same shape:
{
"mcpServers": {
"tinycommand": {
"url": "https://mcp.tinycommand.com/v1",
"headers": {
"Authorization": "Bearer YOUR_KEY_HERE"
}
}
}
}
Restart the host so it picks up the new server.
In your AI chat, try:
"Tell me about my Tiny Command workspace."
The host calls describe_my_workspace and returns your connected integrations, asset counts, and credit balance. From there, ask it to build something or operate your data.
Verify the connection
The wizard's Test connection button calls the server's health endpoint with your key. You can also check it directly:
curl https://mcp.tinycommand.com/v1/health
{ "status": "ok", "protocol_version": "2025-06-18", "build": "..." }
If your host shows the Tiny Command tools but calls fail with an auth error, regenerate the key; the secret is only shown once and a mistyped paste is the most common cause.
Next steps
- Authentication & scopes: choose the right permissions for each key.
- Tool reference: everything your AI can do.
- Security: what data leaves your workspace, and how to lock a key down.