Projects

Set the Frontend URL for browser tasks and the Backend URL for API tasks under General settings in the project's Settings, then click Save Changes. Use staging or another environment you can reset.

Our browsers and requests start from our servers, so both URLs must be reachable from them.

Add instructions and requirements on the Context page. Instructions are rules the agent follows, such as which store to use or what it must never delete. Requirements describe how the product should behave: user stories, specs or API documentation.

We index both, and the agent looks up what is relevant to each task. To add a document, click Add Context under Documentation context in Settings, then paste text, upload a file or enter a URL.

The agent saves notes while it works: test accounts, URLs, flows and decisions you made together. Every task in the project can recall them.

Review the notes under Project memory in Settings. Delete one with its trash icon, or click Clear memory to delete them all.

Put test credentials in the task, or tell the agent to remember them and it saves them to project memory.

If you would rather not share a password, ask the agent to hand you the browser. The live browser opens full screen with the agent's reason at the top. Sign in yourself, then click Resume agent and the agent carries on.

Custom reporting instructions under Agent behavior in Settings change what reports focus on and how they read.

On paid plans, Model defaults set the model for UI, API, Slack and Code tasks in the project. Leave one on Platform default and we choose. See Models.