Connect an AI assistant to Playrunner Cloud

Add the hosted server

In your assistant's MCP settings, add a remote server with these values. The exact field names depend on the client.

Setting	Value
Name	Playrunner
Server URL	https://playrunner.cloud/mcp
Transport	Streamable HTTP
Authentication	OAuth through Playrunner Cloud

1. Start the client's connection or authorization flow.
2. Sign in to Playrunner Cloud, or create an account, in the browser page it opens.
3. Review and approve the requested access.
4. Return to the assistant and ask it to check your Playrunner account.
5. Complete any account setup through the returned browser link, then ask the assistant to check again.

The assistant should call get_account first. Once setup is complete, ask it to list your projects and workflows to confirm that the connection works.

Cloud MCP uses the browser authorization flow. A machine API token from Settings → API tokens is for the CLI or standalone MCP server; it does not replace Cloud OAuth.

Gemini CLI

Merge this entry into mcpServers in your user-level ~/.gemini/settings.json, preserving your other settings:

{
  "mcpServers": {
    "playrunner": {
      "httpUrl": "https://playrunner.cloud/mcp",
      "oauth": {
        "enabled": true,
        "redirectUri": "http://127.0.0.1:7777/oauth/callback"
      }
    }
  }
}

Restart Gemini CLI, then enter /mcp auth playrunner in its interactive prompt. Complete the browser sign-in and consent flow. Use /mcp to inspect the connection, then ask the agent to call get_account.

The explicit callback uses an IP loopback address accepted by Playrunner Cloud; Gemini CLI's default localhost callback is not accepted. If port 7777 is busy, choose another unused port in redirectUri. Run authorization on a computer with browser access. Tool confirmations remain enabled by default.

This setup is for Gemini CLI, not the consumer Gemini chat application. See the official Gemini CLI MCP guide for OAuth discovery and configuration options.

Google Antigravity

In Antigravity IDE, open MCP Servers → Manage MCP Servers → View raw config. Merge the following entry into the existing mcpServers object:

{
  "mcpServers": {
    "playrunner": {
      "serverUrl": "https://playrunner.cloud/mcp"
    }
  }
}

Refresh the server list and use Authenticate for Playrunner in Agent Settings → Customizations. Complete the browser authorization and follow Antigravity's return-to-client instructions. Keep authorization codes in that authentication flow, not in an agent conversation. Then ask the agent to call get_account.

Antigravity documents remote Streamable HTTP and automatic OAuth with dynamic client registration, which Playrunner Cloud provides. A manually configured client secret or Google Application Default Credentials is not needed for this connection. The exact settings location varies between Antigravity surfaces; see Google's Antigravity MCP guide.

Connection verification

These Gemini CLI and Antigravity examples are based on their documented MCP capabilities and Playrunner's Cloud protocol implementation. They have not yet been verified through a complete sign-in and workflow run in those clients. Confirm get_account, workflow discovery, and a test run before relying on a new connection. MCP support alone does not guarantee every client version or authentication flow is compatible.

Access permissions

Permission	Allows the assistant to
playrunner:read	Inspect account and resource metadata, workflows, and execution results.
playrunner:write	Create projects and create or update environments, profiles, and workflow graphs.
playrunner:run	Start saved Cloud workflows using your account allowance.
playrunner:authenticate	Request browser sign-in on a paired companion device.

If a tool reports a missing permission, reconnect through the client's authorization flow and review the requested access. Shared workflows may be read-only; workflow editing requires ownership.

Connect GitHub

Ask the assistant to connect GitHub. The connect_github tool returns a Playrunner browser page where you authorize GitHub and select repositories. Afterward, the assistant can list accessible repositories and branches to configure a workflow. Playrunner Cloud provides the GitHub connection flow.

Tests that need a browser sign-in

An Authentication Profile describes the sign-in page and the condition that confirms a successful login. Creating a profile does not capture a session.

1. Pair the CLI authentication companion on the computer where you will sign in, and keep it connected.
2. Ask the assistant to list your companion devices and Authentication Profiles.
3. Ask it to authenticate the intended profile on the online device.
4. Complete sign-in and any MFA in the browser, then follow the companion's prompt to finish capture.
5. The assistant checks the authentication session and profile readiness before running the workflow.

Enter passwords and MFA codes in the browser. Enter environment secrets in Playrunner Cloud's environment settings. MCP environment configuration accepts non-secret variables and preserves existing secret variables; profile tools return readiness metadata without exposing cookies or browser storage.

Run tests on your own compute

Connect a worker on your laptop, server or VM using Run Cloud tests on your own machine. Then ask your assistant to list local workers and run the saved workflow on the machine you choose. Cloud keeps the execution history, logs and reports. The local execution tool sequence requires an explicit device choice and never falls back to hosted compute.

Troubleshooting

Symptom	What to check
The client only offers a local command or SSE URL	Use a client with remote Streamable HTTP and OAuth support.
The connection asks you to sign in again	Reconnect your Cloud account through the client's authorization flow.
Account setup is incomplete	Open the setup URL returned by get_account, complete setup, then retry.
A tool asks for an additional permission	Reauthorize with the scope named in the error.
A workflow will not start	Check account allowance, access, and that the workflow uses the Playrunner Cloud runner.
No device is online for authentication	Start the paired companion on the computer where the browser will open.
Opening /mcp in a browser shows an error	It is a protocol endpoint. Add it through an MCP client to authorize and call tools.

See Cloud tools for the discovery, authoring, and execution sequence.