People and the agent see the same Space
A Space holds an interface, data, files, logic, and executable services. The agent reads that live environment, not only a note pasted into a chat. What you see in the render view is what it inspects and changes.
Say what should change in a session
Open a session in the current Space. Continue a conversation, start a new chat, or create a new task. State the outcome, such as a source to add, an action to add, or a job that should run on a schedule.
The agent can read the Space, modify it, and run the function. The change stays in that Space, so the next session continues from the same state.
Verify in the render view
After a change, open the render view and use it the way you actually work. If it fails, open the logs and point to the difference in the session. If you need to edit the files that make up the Space, open dev mode.
You can keep more than one Space. Sessions can be renamed or archived. Keep one job in one Space so unrelated work does not share a session.
FAQ
Can the agent chat without changing the Space?
Yes. Ask it to modify the Space, then check the render view, when the result needs to stay. A reply alone does not become part of the workspace.
Where do I look if a change goes wrong?
Check whether the render view still works, then open the logs. Dev mode shows the MCL view, MCS logic, data storage, and files.
How is this different from a normal chat tool?
A chat tool mainly stores the conversation. Here, changes land in the interface, data, files, and executable services. See the comparison guide for the fuller distinction.
Next step
Open an existing Space, create a task, ask the agent to change one thing you will use immediately, and check it in the render view.
Internal links
- How to develop an AI workspace
- How to create an AI Agent Space
- AI Agent Space vs chat
- How to manage multiple AI workspaces