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

# Channels

> Choose where the agent runs, which AI setting powers it, and how it talks — response modes, language, and media.

## What is this?

**Channels** decides where this agent is allowed to work, and configures the finer points of how it communicates once it's there.

\[SCREENSHOT: ai-agent-channels-tab — The Channels tab showing the channel selector pills]

## What can I do here?

* **Communication Channels** — **All Channels**, or pick individually from **WhatsApp**, **Facebook**, **Instagram**, **Web Chat**.
* **AI Setting Source** — which AI setting (provider and API key) powers this agent. If none are listed, add one first under [Settings → AI Setting](/workspace/settings).
* **Allowed Response Modes** — **Text** is always on; turn on **Buttons** and **List** to let it reply with tappable options.
* **Response Language** — **Auto** (mirror whatever language the customer writes in) or **Fixed** (always reply in one language you choose).
* **Media Capabilities** — turn on **Image Understanding**, **Voice Input**, **Voice Output**, and **Video Understanding** individually, for the media types this agent should be able to handle.

## How to use it

### Choose channels deliberately, not just "All"

If you're running separate agents per job (a Sales agent and a Support agent, say), scope each one to the channels where that job actually happens rather than defaulting every agent to **All Channels** — it keeps [multi-agent routing](/workspace/ai-agents/tabs/general#write-the-description-like-it-matters-because-it-does) cleaner.

### Response Language: Auto is almost always right

**Auto** detects and mirrors the customer's own language message by message — the right default for most businesses with customers writing in more than one language. Use **Fixed** only when you deliberately want every reply in one language regardless of what the customer writes.

### Turn on media capabilities only if you actually use them

Each one — Image, Voice Input, Voice Output, Video — is a separate switch. If your Knowledge and Capabilities don't do anything useful with, say, an incoming photo, leave Image Understanding off rather than enabling every option by default.

## Troubleshooting / Technical Notes

* **The builder says there are no AI settings.** Add and activate one under [Settings → AI Setting](/workspace/settings) first.
* **The agent doesn't reply at all.** Confirm this tab includes the channel the message came in on — this is one of the three checks in [Behavior's troubleshooting](/workspace/ai-agents/tabs/behavior#troubleshooting--technical-notes).
* **It replies in the wrong language.** Check whether **Response Language** is set to **Fixed** — that overrides auto-detection on purpose.

## Related docs

* [Behavior](/workspace/ai-agents/tabs/behavior)
* [Settings — AI Setting](/workspace/settings)
* [Create Your First AI Agent](/workspace/ai-agents/create-your-first-agent)