# Bridged Channels
Source: https://dorkos.ai/docs/guides/bridged-channels

Turn a connected Telegram chat into a DorkOS channel, so its whole history lives in one shared log your agent reads before it answers





# Bridged Channels [#bridged-channels]

A chat you connect to an agent is normally a private line: a message goes to one agent, that agent
answers, and nothing about it is visible anywhere else. Start a second conversation with the same
agent, or ask a second agent to help, and the two know nothing about each other.

Bridging changes that. When you bridge a chat, it becomes a **channel** in DorkOS. What people say
in the chat lands in a shared log, and your agent reads that log before it answers. You can also
speak into the chat from the cockpit, and any of the agent's conversations can pick up where the
others left off, because they all read the same history.

<Callout type="info">
  Bridging is being rolled out for Telegram. You turn it on for one chat at a time from the
  **Messaging** half of the [Connections](/docs/connections) page. There is no global switch: a chat
  you do not bridge behaves exactly as it does today.
</Callout>

## How to bridge a chat [#how-to-bridge-a-chat]

1. Open the [Connections](/docs/connections) page and go to **Messaging**.
2. Pick the connected chat you want to bridge, and open its details.
3. Choose **Bridge to a channel**. DorkOS creates the channel and takes you to it.

To stop, open the channel's binding again and choose **Un-bridge**. The chat goes back to a private
line, and the channel is archived, not deleted, so its history is kept.

## Three things to know before you bridge [#three-things-to-know-before-you-bridge]

Bridging a chat is a real change in who can reach your agent. Read these first.

### People you may not know can put text in front of your agent [#people-you-may-not-know-can-put-text-in-front-of-your-agent]

A chat can have other people in it, and a bot in a public chat can be found by anyone. Once a chat
is bridged, what those people type becomes part of the log your agent reads. Your agent treats their
words as messages, never as instructions, and every outside person is clearly marked with the
platform they came from. But the words are still theirs, so bridge a chat only when you are
comfortable with who can write into it.

### The permission mode is the real bound [#the-permission-mode-is-the-real-bound]

A bridged chat starts in the mode that **asks you before your agent does anything sensitive**. That
prompt is your protection. Raising the mode, so the agent acts without asking, is a decision you are
making with a stranger possibly on the other end of the chat. DorkOS never raises it for you.

### The room log is the audit trail [#the-room-log-is-the-audit-trail]

Every message that reached your agent is in the channel's log: durable, kept in full, and never
trimmed. That is a real gain over a private line, where the same message reached the agent and left
no record you own. If you ever want to know what your agent saw and when, the log is the answer.

## What a bridged channel shows you [#what-a-bridged-channel-shows-you]

* **A visibility badge**, on a bridged group, tells you whether your agent sees every message in the
  chat or only the ones that mention it. This reflects the chat platform's own privacy setting for
  the bot, not a DorkOS setting. A one-to-one chat has no badge, because there is nothing there the
  bot cannot see.
* **Origin marks** sit next to everyone who joined from the outside, showing the platform they are
  on, so you can always tell a person on your own machine apart from someone writing in from a chat.

## What bridging does not do [#what-bridging-does-not-do]

* It does not import old messages. Chat platforms do not hand a bot its history, so a bridged
  channel starts its log from the moment you bridge. If the agent already had a conversation going in
  that chat, it can carry on from there, but the earlier messages stay in that conversation, not the
  channel log.
* It does not make your agent injection-proof. An agent that reads an outside message and acts on it
  can still do things within the permissions you granted. The permission mode above is the bound that
  matters.
