Connect MCP

Connect an AI client to https://api.operio.run/mcp to manage tests, start runs, and inspect results without installing the CLI.

Use a client that supports Streamable HTTP and OAuth. Sign in through the browser; no API key is needed. Approve read access to inspect data, or read and write access to create, edit, run, and delete resources.

Client setup

{
  "mcpServers": {
    "operio": {
      "url": "https://api.operio.run/mcp"
    }
  }
}

Run a test through the agent

Ask the agent to list projects and use the returned project ID. Then choose published tests and one target: an environment for web or a build for mobile.

Run the Smoke suite against Staging. Wait for it to finish, then summarize the failed checks and their evidence.

No test case for it yet? Ask for an exploratory run instead, and the agent pursues a plain-English goal without test cases:

Start an exploratory run against Staging with the goal "Buy the cheapest item as a guest". Wait for it to finish and report what it found.

Runs are asynchronous and use credits. A finished run can contain failed cases, so the agent must inspect the results.

Upload mobile binaries through the web app or CLI first. The hosted server cannot read local files.

Access and accounts

The connection uses the account approved in the browser, independently of your CLI login. To switch accounts, disconnect and authorize again as the other account.

Each call checks your current team membership. Stored login passwords are never returned by tools, though write access allows the client to create and update test accounts. Write access also allows changes and deletions, as well as runs against your apps.

Troubleshooting

See coding agent workflows for test creation and failure investigation, or CLI setup for a local connection.