The Jollop agent

Jollop has a built-in agent, think of it as a deploy-native teammate. You tell it what you want in plain language and it does the work by driving the platform’s own API.

What it does

• Builds and deploys. “Build a poll app and deploy it.” It writes the app’s source, deploys it, and returns the live URL.
• Verifies its own work. After deploying it checks the app actually serves, reads logs, inspects the run contract, and fixes problems itself before reporting success.
• Manages deploys. “Make quick-polls public.” “Show me the logs for the api.” “Delete the devon-tower deploy.” It lists, shares, inspects, edits contracts, and (with explicit confirmation) destroys.

How it works

The agent runs a tool-use loop, but every tool is a call to the platform’s real API, so it exercises the exact same pipeline the CLI and the browser do. It adds no separate deploy path.

It acts as you

The agent runs as the identity that invoked it. When you use it signed in, its actions are authorized exactly as if you did them by hand, it can only touch apps you can touch. It cannot be used to reach across tenants.

Safety

Destroying an app is permanent, so the agent will only do it when you have explicitly asked to delete that specific app, and it streams a clear line before it happens. Vague cleanup requests get a question, not a teardown.

Use it from the Agent panel in the app, or through the MCP server’s agent tool from any MCP-capable client.