# Manual actions > Manual actions are steps a person has to do outside Harny. The agent hands them off, you mark them complete, and the agent is told. Some steps cannot be done by the agent. A call, a signature, a grant of access, or a judgment call belongs to a person. The agent hands that step off to you. You do it outside Harny, then mark it complete. ## What a manual action is A manual action is a step a person must do outside Harny. Examples are a phone call, a signature, giving access to a system, or a judgment call. The agent does not perform these steps. It creates a handoff that describes what needs doing. You decide when it is finished. ## The lifecycle 1. The agent hands off. It creates a manual action with `POST /v1/handoffs`. 2. You do the step outside Harny. 3. You mark the action complete. This calls `POST /v1/handoffs/{id}/resolve`. 4. The agent is told that the step is complete. Marking an action complete does not do the step a second time. It also does not create a rule. The next time a similar case comes up, the agent hands off again unless your action settings change. ## Where to find them In the console, open **Actions**, then **Manual actions**. You reach Automation from the Manual actions page. ## API routes | Route | Purpose | |---|---| | `POST /v1/handoffs` | Create a manual action (the agent does this) | | `GET /v1/handoffs` | List manual actions | | `GET /v1/handoffs/{id}` | Read one manual action | | `POST /v1/handoffs/{id}/resolve` | Mark a manual action complete | ## What completing one does not do - It does not repeat the step. You already did it. - It does not create a rule. Future actions follow the settings that already exist. Harny can offer a change to an action kind's setting when recorded responses support it. You apply the offer or keep the current setting. See [Learning](/docs/agent-harness/learning) for how offers work. ## Next - [Permissions](/docs/agent-harness/permissions) for the three modes per kind of action - [Learning](/docs/agent-harness/learning) for offers to change an action setting - [Overview](/docs/agent-harness/overview) for how the agent loop works