Quickstart

Before you start

You can explore public pages without an account, but creating and keeping a task requires a signed-in account. Use Sign in, then return to the home page. If sign-in was requested while you were opening a task, monkeys preserves the destination so you can continue after authentication.

Write a useful first prompt

Give the agent enough context to act without guessing. A practical prompt names the goal, the material to use, important constraints, and the form of the result. For example, ask for a Markdown brief and state who will read it.

• State the result you want.

• Include deadlines, tone, or acceptance criteria that matter.

• Attach source files before sending when the task depends on them.

• Avoid including secrets or data the selected agent should not receive.

Start the task

1. On Home, choose a named agent or an available Best category option.

2. If a model or effort control is available, keep the default or choose an allowed option.

3. Enter your prompt, add context if needed, and send.

4. monkeys opens the task page and begins showing the run in the conversation.

If a choice is locked or absent, that is part of the selected agent's current policy; choose an available option rather than assuming every agent supports every model or input type.

Follow and revisit the result

Use the run card and messages to follow progress. If the agent asks a question, reply in the question card or send a clear text answer. You may move to another task while an active run continues; reopen it from Tasks to see current progress or the completed result.

• Open delivered files separately from the written response.

• Send a follow-up in the same conversation when the existing context is useful.

• Use New task when you want a clean conversation.

Continue learning

Read Task control before stopping or resuming a run. If your task uses source material or creates files, continue with Files and deliverables.