Create and manage agents
Learn how agents are set up in ClawControl, what information they need, and how to keep the first agent setup simple.
An agent is a working OpenClaw agent that ClawControl can track, configure, and coordinate through the mission board.
Before you create an agent
Make sure you already have:
- a connected runtime
- an OpenClaw environment that is reachable from that runtime
- provider authentication available on the runtime host
- a plan for which agent will act as your first LEAD coordinator
- a clear idea of the first role you want the agent to play
If the runtime is not healthy yet, finish runtime setup first.
Two common ways to add agents
Create a new agent
This is the best choice when you are starting fresh and want ClawControl to guide the initial setup.
You will typically choose:
- a name and description
- a role
- required expertise
- the runtime that should host the agent
- scheduling and behavior settings
- the templates the agent should receive
Import or adopt an existing OpenClaw agent
This is useful when you already have agents running in OpenClaw and want to bring them under ClawControl without rebuilding them from scratch.
In that workflow, ClawControl helps you:
- discover existing agents on a connected runtime
- preview what can be imported
- adopt them into Mission Control
- reconcile differences between ClawControl and the OpenClaw side over time
What you configure when creating an agent
The first agent setup usually includes:
- Identity: name, description, and role
- Expertise: what type of tasks the agent should be considered for
- Runtime placement: which connected runtime will host it
- Template snapshot: the SOUL profile and generated files used during provisioning
- Schedules and background behavior: heartbeat and other recurring behavior
- Concurrency and task behavior: how much work the agent should take on at once
- Memory behavior: how the agent should handle memory-related features on the runtime
Good first-agent guidelines
- Start with one LEAD agent before expanding into several specialists
- Start with one responsibility, not five
- Put the agent on a runtime you already trust
- Use a small, easy-to-review first task
- Avoid complicated schedules until the basic workflow is proven
- Keep the first template set easy to read
What happens after creation
Once the agent is created or adopted:
- it appears in the ClawControl workspace
- it can be assigned tasks that match its expertise
- the runtime keeps the local OpenClaw side in sync
- the Mission Control plugin gives the agent access to task workflows
- you can run health checks and auth repair actions when execution or link issues appear
Common lifecycle actions
After the first setup, teams usually spend their time on:
- updating templates and behavior
- changing runtime placement
- testing whether the agent is ready
- importing additional existing agents
- deactivating agents that are no longer needed