> ## Documentation Index
> Fetch the complete documentation index at: https://campus.agile-academy.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Agents
> Build your own personal AI agents — give them a routine, connect your tools, and get results by email or Slack.
**Agents** is a playground for building your own personal AI assistants inside Campus. An agent is a small worker you describe in plain language — "research the latest AI news every morning and write me a short digest" — that can run on a daily routine, use the tools you connect, and deliver its results back to you.
You find it under **Agents** in the main navigation.
Agents is currently in **beta** and available by invitation. If you do not see it in your navigation, it is not yet enabled for your account — ask your trainer or contact support.
## Creating an agent
There are two ways to build an agent:
Describe what you want in a chat. Henrik asks a few questions, then sets the agent up for you — including a routine and suggested tools.
A three-step wizard: **Describe** what the agent should do, give it a **Schedule** (optional), then **Connect** channels and services.
If you are not sure where to start, pick one of the starter examples — **AI news scout**, **Retro facilitator**, **Notes summariser**, or **Daily briefing** — and adjust the name and instructions from there.
The instructions are the heart of the agent. Be specific about the task, the inputs it should use, and how it should deliver results. You can change the name and instructions later in the agent's **Settings**, or click **Edit with Henrik** and describe the change in chat.
## Routines
A **routine** is a daily run time plus a task — the instruction your agent receives each time it wakes up. An agent can have no routine, one, or several.
* Routines run **once per day** at the time you pick (presets at 09:00, 13:00, and 18:00, or any custom time). Times use your local timezone, detected automatically.
* If you leave the routine's task empty, the agent simply follows its general instructions.
* Each routine has a **Run on schedule** toggle. Pausing it stops automatic runs — you can still trigger the agent yourself.
* An agent without any routine only runs when you trigger it: with **Run now**, by email, or via Slack.
During the beta, routines are daily — an agent runs at most once per day per routine.
## Running and steering
The **Runs** tab is where you work with your agent directly:
* **Run now** starts a run with the agent's usual task; typing a one-off instruction starts a run with that instead.
* You can **watch a run live**, send follow-up messages mid-run to steer or correct the agent, and **interrupt** a run that is going the wrong way.
* Every run shows its status — Queued, Running, Done, or Failed — and past runs stay available in the run history, together with the files each run produced.
## Channels
Channels control how the agent communicates with you. Both email directions are separate toggles under **Apps & tools**:
* **Agent can email you** — the agent can send reports and results to your Campus email address (and only to you).
* **You can email the agent** — the agent gets its own private email address. Emailing it starts a run: your subject and message become the task, and attachments land in the agent's Files workspace. Keep the address to yourself.
* **Slack** — connect a dedicated Slack app (Campus generates the app manifest for you; you create the app in your workspace and install it). Once connected, the agent can read and post in Slack, replies in-thread, and answers when mentioned or DMed. You can optionally let it respond to all messages in its channels instead of only @mentions.
* **Telegram** is coming soon.
## Connections
Connections are the external services your agent can read and write **on your behalf** — for context ("read my Notion notes") and actions ("create a Linear issue"). You authorize each service individually: click it, sign in on the service's own page, and you are back in Campus. You can disconnect any service at any time.
Suggested connectors include **Notion, Linear, Miro, Figma, and Canva**, with many more across project management, docs, dev tools, data, CRM, automation, finance, and research — use the search in **Apps & tools** to find them.
**GitHub** works slightly differently: you connect a specific repository with a fine-grained personal access token (Contents and Pull requests scopes). The agent can then read the code, edit it, commit, open pull requests, and manage issues — it works on a branch and opens a PR rather than pushing to your default branch.
## Files
Every agent has a persistent **Files** workspace that carries across runs, grouped into:
* **You added** — files you upload; available to the agent in every run.
* **Received** — attachments that arrived by email or Slack.
* **Made by the agent** — the reports and artifacts your agent produced.
Everything is downloadable, and files are never silently overwritten — a new version of `report.html` becomes `report-2.html`.
## Memory
Agents remember things across runs — preferences you stated, context they gathered, decisions you made together. The **Memory** tab shows exactly what an agent has remembered, and you can review, edit, add, or remove entries yourself.
## Agents and Henrik
Henrik and your agents work together:
* **Create with Henrik** builds a new agent from a conversation.
* **Edit with Henrik** changes an existing one the same way.
* In any Henrik chat, you can **hand a task off to one of your agents** — Henrik dispatches it, the agent works in the background, and the results appear on the agent's page in Campus.
## Privacy, limits, and deletion
* **Agents are private to you.** Only you can see your agents, their runs, their files, and their memory.
* Connections act with the access **you** granted — an agent never touches accounts you did not connect.
* Routines run at most once per day per routine during the beta.
* Deleting an agent (Settings → Danger zone) removes the agent along with its schedule, channels, and connections. Run history is kept but hidden. This cannot be undone.