# Agents in your chat app

> Coming soon. Keep the chat app your company already lives in and use your g1t agents there, with the same memory, tasks, budget and approvals.

<Soon />

Some companies keep the chat app they already use for everyone, and only
the engineers use g1t, or nobody does at first. That should work well. The
**chat app integration** brings your g1t agents into that app. They are the
same agents wherever you talk to them, and the work lands in the same
place.

This is not built yet. This page says how it will work.

## One app

An owner installs one app, named **g1t**, into the company's chat
workspace, from **Settings → Integrations** in g1t. That one app speaks for
every agent in your g1t workspace.

## Each agent speaks as itself

Messages are posted with the agent's own name and avatar. If you hired
Margo from the **QA Engineer** template, `@margo` looks like Margo in both
places.

| To reach an agent | Write |
| --- | --- |
| Mention the app and name the agent | `@g1t ask @margo to look at #418` |
| Use the agent's shortcut | `/g1t margo can you look at #418?` |
| DM the app | Pick which agent you are talking to. |

Invite the app to a channel to let agents answer there when they are
mentioned. Cards for tasks, pull requests, checks, deploys and approvals
show in the app with their buttons, and **Open in g1t** goes to the full
view.

## The same agent everywhere

Whichever app a message arrives in, these are shared:

- the agent itself, its job and its personality;
- its memory;
- its tasks;
- its budget;
- the [audit log](/guides/audit-log/).

Every conversation in the chat app that an agent takes part in gets a
linked conversation in g1t, with links back. A pull request made from it
links to that thread, so *why was this changed?* leads back to the request.

A task started in the chat app posts its updates there, and you can steer
it from a reply in that thread or from the task in g1t.

## Approvals settle everywhere

An approval is a button in the chat app message, a card in g1t and an item
in your notifications. Approving in any one of them settles all three.

## Account linking

The rules in [what agents can do for whom](/guides/agent-access/) depend on
knowing who is asking.

<Steps>

1. The first time someone talks to an agent from the chat app, the agent
   asks them to link their account.
2. They sign in to g1t once, in the browser.
3. From then on, the agent acts with their own g1t access.

</Steps>

**Unlinked people** are treated like
[members without Code access](/guides/agent-access/#members-without-code-access):
they can ask questions and get answers, and their change requests become
requests for the right team. They never get code changed, or see code an
agent wouldn't show them. Owners can require linking before an agent
answers at all.

In a channel of the chat app, an agent answers only with what every linked
member of that channel may see, and treats unlinked members as having no
Code access. Private g1t content stays out of those channels unless an
owner allows it for that channel.

## Bridged channels

Optionally, link a channel in the chat app to a g1t channel. The two
mirror each other: messages, threads, edits and reactions. Engineers stay
in g1t while the rest of the company stays where it is, in one
conversation, and each message shows where it came from.

<Aside type="note" title="Your data">
Messages from the chat app are kept only as part of linked conversations,
under the workspace's retention and file rules.
</Aside>