Introducing Agent Mode: set a task and walk away

We've been building toward this for months. Today, Agent Mode is live for all Pro and Max plan users. It's the biggest feature we've shipped since launch.

What is Agent Mode?

Agent Mode lets you describe a complex task in plain English and have an AI agent handle the entire workflow: planning, coding, testing, and iterating. You approve the plan, then walk away. The agent builds while you do something else.

How it works

The workflow is four steps:

1. Describe it -- Tell the agent what you want in plain English.
2. Plan it -- The agent breaks your goal into clear steps with estimated timelines.
3. Approve it -- Review the plan. Edit, reorder, or reject anything.
4. Build it -- The agent writes code, creates files, runs tests, and fixes issues.

Self-healing agents

When something breaks, the agent diagnoses the issue, applies a fix, and keeps going. You only get asked when human input is truly needed. Errors become speed bumps, not roadblocks.

The best agent interaction is one where you forget it's running, then come back to find everything done.

Availability

Agent Mode is available now on Pro ($60/mo) and all Max plans. Starter plans use standard chat mode.

Get started today by downloading Mulu Code or upgrading to Pro in your dashboard.