# Auto mode

> AI agents: read the [complete documentation index](https://chiefisnext.com/llms.txt). Append `.md` to any docs URL for its Markdown version.

Auto mode lets Chief approve and dispatch its own proposed tasks within the limits you choose. Risky or costly tasks are held for you. It works toward your existing [goals](https://chiefisnext.com/docs/goals); read a few [milestone results](https://chiefisnext.com/docs/milestones) before enabling it on a project.

## Before enabling

Connect the relay, check agent readiness in **Tutorial**, and give Chief a clear goal and constraints. Keep the relay running and the computer awake. Your provider's billing and Chief's plan limits still apply.

## Enable a bounded run

1. Select **Auto** in the project header to open **Auto mode**. You can also find it in project settings.
2. Choose a ready **Worker agent**. Every auto-mode task uses this agent for the duration of the run. The choices include Claude Code, OpenAI Codex, Qwen Code and Kimi Code, plus the project's current default when it is another provider.
3. Set **Max tasks**. A blank value means no task limit; start with a small number such as `3`.
4. If available for that worker, set **Weekly usage cap** and **5-hour usage cap**. A blank field means no cap.
5. Read the capability message and select **Enable**.

**Expected result:** the project shows an active auto-mode banner with task counts and any measured usage. Chief evaluates proposals and dispatches eligible work as dependencies permit.

## Understand usage caps

Caps measure percentage points of the provider's usage window consumed since enabling or resuming. A cap of `10` means ten percentage points of additional window usage, rather than ten dollars or a ten-percent account balance.

Usage is account-wide: it includes Chief's planning calls and other use of the same provider account. These are soft limits checked before dispatch. Tasks already running may finish and add more usage.

Some providers do not report one or both usage windows. Disabled inputs and the capability message identify the unavailable caps. When no usage reporting is available, Chief cannot measure or enforce a usage budget; only a task limit you set applies. Do not assume every selectable agent supports the same caps.

## Review held tasks

A **Held by auto mode** badge identifies work left for you. Open it to read the reason, estimate and cost drivers when provided. Review the proposal and use its normal approval or revision controls. A held proposal needs your judgment; it is different from a failed run.

Check [manager tasks](https://chiefisnext.com/docs/manager-tasks) for decisions that block work. Auto mode does not clear those requests for you.

## Resume or disable

When auto mode pauses, read the banner's reason. Select **Resume**, review the limits and use **Resume auto mode**. Resuming starts a new round: you can raise, lower, add or remove any limit, and the task and budget counts restart from zero under the new limits. While a run is active, limits can only be raised or removed.

Use **Disable** in the auto-mode settings to end the auto-approval run. Review running tasks separately; disabling auto mode controls further automatic approvals rather than serving as a task-cancellation action.

The project's **Agents** selector is locked while an auto-mode run is active, including when paused. To [switch the worker](https://chiefisnext.com/docs/agents#switch-the-project-providers), disable auto mode, choose the intended provider and enable a new run with appropriate limits.

## When progress stops

Check the pause reason, held proposals, manager tasks and relay status. Follow [Troubleshooting](https://chiefisnext.com/docs/troubleshooting) for provider failures or usage limits. Use [Schedules](https://chiefisnext.com/docs/schedules) for work authorized to run at a specified time or repeating cadence.