> ## Documentation Index
> Fetch the complete documentation index at: https://lab.pollack.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Agento University
> A visual operating environment for a fleet of agents that each own a codebase
IN DEVELOPMENT
Exploratory — product vision
Give every repository an agent that owns it. Then give yourself somewhere to watch them all.
Agento University is the graphical surface for the
steward pattern — a workspace for the agent that owns one codebase, and a campus view for the fleet that owns the rest.
## The Problem It Solves
The [Forge methodology](/methodology/forge) is a command-line product. You run a slash command, an agent scaffolds a vision or works a roadmap step, and the result is files in git. That works, and it is how the methodology is used today.
It does not scale to *watching*. Once a dozen repositories each have an agent continuously accountable for them, the questions stop being "what did this session do" and start being: which of these needs me right now? What is that one waiting on? What did it change while I was gone? A terminal answers those one repository at a time.
Agento University is the answer to the second question. It is not a different methodology — it is a **higher-level interface onto the same stewards**, aimed at the point where a fleet becomes too large to hold in your head.
## Two Surfaces
### The Studio — working with one agent
The Studio is the workspace for a single steward. It puts four things in one shell:
* **The workflow canvas** — a rendered view of the agent's workflows. Clicking a node opens the exact line of source that declared it, so the picture and the code never drift apart.
* **The artifact panel** — the agent's documents and files, with a viewer that opens at a line and highlights it.
* **The inbox** — the agent's message lanes, so obligations arriving from elsewhere have a human end.
* **The action bar** — run a turn; or forge a new steward into existence and start its first conversation, without leaving the shell.
### The Campus — watching the fleet
The campus is the fleet view: each repository is a building, each building carries a live status billboard, and the state you are looking at is written by the stewards themselves rather than assembled by a separate monitoring service.
The metaphor is a navigation aid, not a theory. A building is a repository; a billboard is that repository's current state. Where the metaphor stops being useful, the interface uses ordinary words.
SCREENSHOTS PENDING
Captures of the Studio shell, the workflow canvas, the inbox, and the campus billboards are being prepared from the running application.
## What Is Actually Built
The shell is real and tested; the product thesis on top of it is not yet proved. Stating that boundary is more useful than a launch date.
| Surface | State |
| ---------------------------------------------------------------------- | --------------------- |
| Studio shell — sidebar, advisor tabs, artifact panel | **Built** |
| Workflow canvas with click-through to declaring source | **Built** |
| Inbox — message lanes with a human end | **Built** (read-only) |
| Action bar — run a turn, forge a new steward | **Built** |
| Campus — buildings, billboards, interiors | **Built** |
| Review-loop cockpit — the product thesis | In progress |
| Runtime surfaces — live run watching, approvals, experiment dashboards | Deferred |
## On "Graduation"
The campus metaphor supplies a word that turns out to carry a real engineering meaning, so it is worth stating plainly rather than leaving as a pun.
**An agent graduates when it reliably meets a repeatable benchmark.** Not when it produces one good result — when the judges that scored its development converge, stop being development scaffolding, and become a measuring stick it clears consistently. That is a threshold you can compute, and it is the same threshold the [four-tier jury](/methodology/four-tier-jury) exists to measure.
Everything else about the campus is navigation. This part is not.
## Where It Sits
Agento University is **exploratory and long-term**. It is deliberately not part of the control-loop architecture: the [Improvement Flywheel](/methodology/improvement-flywheel) measures and steers a single agent, and it does that whether or not any of this exists. The University is a layer above — coordination and visibility across many agents, once the single-agent problem is solved well enough to have many.
The projects it is built on — [Agent Workflow](/projects/agent-workflow) as the engine, [Agent Journal](/projects/agent-journal) for traces, [Agent Judge](/projects/agent-judge) for evaluation — are shipped and documented today. Agento University is the product surface being assembled on top of them.
## Related
What a steward is, and what the University is a fleet view of
The methodology and the command-line product underneath