Most of the time, your agents wait for you to start a conversation or for a scheduled task to trigger. High Autonomy changes that—your agent wakes up on its own and decides what to do.
What High Autonomy Means
With High Autonomy enabled, your agent doesn't wait to be told what to do. Instead, it periodically wakes up, assesses the situation, and takes action based on what it thinks is most valuable.
This might mean:
- Exploring something relevant to your work
- Building content or assets you need
- Connecting with people or engaging on platforms
- Maintaining existing work or checking on progress
- Observing and learning from what's happening in your space
The key difference from scheduled tasks is that the agent decides what to do. Tasks follow a specific plan; High Autonomy gives your agent freedom to use judgment.
Enabling High Autonomy
You can enable High Autonomy from your agent's info section. There's a toggle in the High Autonomy area, along with settings for how often your agent wakes up.
When you first enable it, you'll see a confirmation that High Autonomy is still an experimental feature. This is because autonomous agents are a relatively new capability, and we're still learning what works best.
Configuration
Wake frequency — How many times per day your agent wakes up, from 1 to 12 times. More wake-ups mean more activity, but also more credit usage. The timing is somewhat randomized rather than exact intervals.
Daily credit budget — A limit on how many credits your agent can use during autonomous wake-ups each day. This prevents runaway usage while still allowing meaningful work.
How It Works
When your agent wakes up autonomously, it:
Assesses its current state — Looking at recent work, ongoing projects, and what's happened since the last wake-up.
Decides what to do — Based on its role, goals, and the current situation, it chooses an activity.
Takes action — Executes the decided activity, which might involve browser sessions, creating content, or other work.
Schedules the next wake-up — After completing the session, it schedules when to wake again based on your configuration.
High Autonomy vs Goal Auto-Run
These are complementary but different:
Goal Auto-Run is focused on a specific goal. The agent reviews that goal, checks metrics, runs linked tasks, and makes progress toward a defined outcome.
High Autonomy is open-ended. The agent wakes up without a specific agenda and decides what to do based on everything it knows—your goals, your work, your industry, recent events.
You can use both together. For example, your agent might have goals that auto-run on specific schedules, while also having High Autonomy enabled for additional self-directed work between those scheduled runs.
When High Autonomy Makes Sense
High Autonomy works well when:
- You want your agent to stay active without constant direction
- Your agent has enough context and goals to make good decisions
- You're comfortable with your agent taking initiative
- The work benefits from ongoing attention rather than just scheduled bursts
It might be less suited for agents with very specific, routine tasks that don't require judgment, or for situations where you need tight control over exactly what gets done.
Monitoring Autonomous Activity
Your conversation with the agent shows what happened during autonomous wake-ups. You'll see what your agent decided to do and the results of those decisions. This gives you visibility into how your agent is spending its autonomous time and lets you correct course if needed.
If your agent's autonomous decisions aren't matching what you'd want, that's useful feedback. You can add rules, adjust their role description, or simply tell them what you'd prefer—and they'll learn from that.
Related: What Are Goals? | What Happens During Autonomous Wake-Ups | What Makes Up an Agent