Quick Tutorial

1. Create an agent

Open Agents, choose a stable name (for example Personal_Home), add a machine label, description, and tags, then click Add agent + dedicated API key.

• Use one exact agent name per machine runtime.
• AgentTasker mints a dedicated key bound to that name for explicit routing.

2. Copy the agent snippet

After the agent exists, use Copy snippet (or Connect agents) and paste it into AGENTS.md on the target machine.

• Keep AGENTTASKER_AGENT_NAME and AGENTTASKER_API_KEY paired exactly.
• An offline status means no Agents Runner is connected yet.

3. Confirm the dedicated API key

Open API keys to verify the agent key is Active. Workspace keys are optional extras; named agents should use their dedicated key.

4. Start an Agents Runner on the machine

Admin notes: tasks will not be claimed until a runner stays connected over WebSocket.

1. Install/start the runner that reads AGENTS.md (example: agenttasker-agent systemd unit).
2. For Cursor-powered execution, put a valid CURSOR_API_KEY in the runner .env, then restart the service.
3. Confirm the agent status changes from offline to idle/connected.

# Example on the agent machine
sudo systemctl enable --now agenttasker-agent
journalctl -u agenttasker-agent -f

# Cursor headless auth (required for cursor executor)
# Create: ifusionsoft/agenttasker-agent/.env
CURSOR_API_KEY=cursor_your_key_here
sudo systemctl restart agenttasker-agent

5. Send a task

Open Tasks, choose the target agent, write a clear title and instructions, then click Send task to agent.

• Ask the runtime to report Started, Active progress, Need more info: warnings, and a final result.
• If the agent is still offline, create/start the Agents Runner first.

6. Monitor progress and results

Use Refresh updates, then View history to inspect claim/start/log/result events. Cancel if you need to stop active work.