Have a coding agent run monito check

A coding agent can ask Monito whether the change works. The check is the same one you would run: a URL, a sentence, a verdict.

monito check https://your-app.example/signup "The form rejects a malformed email and explains why" --json

Exit 0 is passed. Exit 1 is failed or error. The JSON carries status, cause, causeLabel, reportUrl, and firstFailure. The agent should read reportUrl rather than guess. Setup for each client is Claude Code, Cursor, and Codex. The overview is AI agent usage.

MCP

Hosted MCP is https://www.monito.dev/api/mcp. No CLI install. The agent calls typed tools. The default tools cover checking, reading evidence, and saving a test. all adds CI setup and the rest of the catalog. Tool names and arguments are the MCP docs.

Hosted MCP cannot open localhost. For a dev server, the agent uses monito check or the local stdio server, monito mcp. Keep that process running for the whole check.

monito check save <checkId> --name "Test name" is how a one-off check becomes a saved test. It does not create a schedule.

CI, if you want it blocking

After monito auth login and gh auth login, the agent can plan, apply, and verify a GitHub Actions workflow. plan is read-only. apply writes the workflow. The job uses GitHub OIDC. It does not print an API key. monito ci verify --remote-auth checks auth and spends no credits. A real dispatched run spends credits, and the CLI asks you to confirm.

The CLI does not commit or push. Fork pull requests are skipped.

What the agent must not be told

Monito does not heal a failing scenario. It does not decide which flows matter. Chromium only. A localhost check cannot be scheduled. Prices are on /pricing.