Agentic AI
Standing instructions that watch, and act.
Agents, their runs, and the studio where they are written. An agent runs under an identity with roles and can do exactly what that identity can do — no more. Every run is recorded with what it looked at and what it changed.
Watch for: Start in propose mode. An agent that acts is only as good as the condition you wrote, and proposals are the cheap place to find that out.
Lessons opened: 0/9
Design
What an agent is, what it may call, and the canvas it is joined up on.
- Build What an agent is, what it may call, and what it is allowed to know. 4 screens
- Orchestrate The canvas, the graphs on it, and what starts them. 4 screens
- Templates Working agents for problems most businesses have. 1 screens
Reach
The systems outside this platform that agents can talk to.
- Connect The MCP servers, APIs and other agentic platforms this company reaches. 3 screens
Authority
The limits, and what broke them.
- Control The limits that are not the agent own, and what broke them. 2 screens
Running
What is happening now, what is waiting on somebody, and what came out.
- Operate What is running, what is waiting on somebody, and what it produced. 5 screens
- Assess Whether any of it was any good, and what the agents remember. 2 screens
Reporting
Cost, value, and whether any of it is trusted yet.
- Reporting What the agents cost, what they produced, and how much of it was kept. 8 screens
More
- Reachable by link These screens carry no tile on the module page. They are reached from the record they belong to, from a saved view, or b 3 screens