> ## Documentation Index
> Fetch the complete documentation index at: https://docs.syvon.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Agents

> An agent is a configured content-maker — one job, one brand, its own project and chat.

# Agents

An **agent** is a configured content-maker. It has a name, a job, a brand,
and a flow — and its own chat and working set. When you "talk to Syvon", you
are always talking to a specific agent.

```text theme={null}
agent
├── identity    name, what it's for, which brand it wears
├── flow        how it makes things (see Flows)
├── chat        the conversation where you direct it
└── project     its files: drafts, renders, published output
```

## One agent per job

The useful unit of separation is the *job*, not the format:

* "Launch teasers" — one agent, punchy short video
* "The weekly digest" — another agent, calmer, longer
* "Product one-pagers" — another, document-shaped

Because each agent keeps its own conversation and project history, a
well-scoped agent gets better at its job over time — and a do-everything
agent just gets muddy.

## Agents are public or private

An agent stays internal until you **publish** it. A published agent gets its
own public surface — the wrapper page where people can meet it, read its
feed, and (if you allow) chat with it. Publishing is the same act that puts
its content on your site (see [Publishing](/learn/publishing)).

## Directing an agent

Everything is the conversation:

* **Briefs** — "make a 15s teaser for X" starts a piece.
* **Corrections** — "swap the second shot for a product close-up" changes it.
* **Standing instructions** — things you say repeatedly belong in the agent's
  setup, not in every brief.

## Related

* [Create an agent](/learn/tutorials/create-an-agent) — the hands-on version
* [Flows](/learn/flows) — the "how it makes things" half