Messaging channels

Channels connect an agent to an external conversation surface. They belong to a profile on a server gateway. The desktop's built-in local gateway disables channels; a desktop connected to a server can configure that server's channels.

Open Customize → Channels for the selected agent. Channel credentials authorize that connection; they do not grant owner administration of the entire gateway.

WhatsAppCopy Anchor Link

Jian supports paired WhatsApp connections and the WhatsApp Business integration. They have different setup requirements.

For pairing, start the connection flow and follow the QR/pairing instructions. Keep the gateway running to receive messages. Inspect connection state if the account disconnects; scanning a QR once does not guarantee it remains connected indefinitely.

A WhatsApp Business connection uses Meta's credentials and webhook configuration. Follow the fields and callback information shown for that connection. Platform rules and messaging restrictions still apply; a paired account and a Business account are not interchangeable credentials.

TelegramCopy Anchor Link

Create a bot through Telegram's bot setup flow and enter its token in the channel dialog. The server must be reachable as required by the bot's webhook setup. Protect the token and rotate it if exposed.

Check the gateway's public URL and HTTPS/proxy behavior when setup succeeds locally but incoming Telegram messages never appear.

Contacts, groups, and AI repliesCopy Anchor Link

A new sender can become a pending contact request. Approve or block them from Jian. Permission to chat and permission to approve actions are separate settings.

An existing contact's AI replies can be paused without treating the channel as disconnected. Groups have their own access and addressing rules. Agents can follow group context without responding to every message; being explicitly addressed matters.

For a protected action, a contact must have approval rights. An Ask first conversation without an eligible approver refuses the action with an explanation. Switching off a contact's AI replies is not the same as revoking the entire channel credential.

Conversations and deliveryCopy Anchor Link

Channel sessions appear alongside the agent's other conversations. The owner can inspect the same work in Jian. Messages, media generation, and outbound delivery are separate stages: a completed model run does not prove the messaging platform delivered its output.

Schedules can target a channel conversation so their result is delivered there. Another conversation can send an explicit message/file to a known session when authorized. Review recipients and approval scope before sending beyond the current conversation.

API channelsCopy Anchor Link

API channels provide a channel token, native ingress, and an OpenAI-compatible chat base URL. They apply contact policy too; they are not an anonymous route around the owner's settings.

Use the HTTP API guide for curl examples, metadata.chat_id, streaming, idempotency, and compatibility limits. The OpenAI-compatible endpoint is text-only; normal native media support should not be inferred from that endpoint.

A diagnostic sequenceCopy Anchor Link

1. Verify the gateway is online and the channel reports connected/configured.
2. Check whether the sender is pending, approved, or blocked.
3. Check whether AI replies are enabled for that conversation/contact.
4. Find the run and inspect provider or tool errors.
5. If work completed, inspect outbound delivery separately.

For native API ingress, a 202 with accepted: false is possible. Inspect the response body before calling the integration successful.

Implementation referenceCopy Anchor Link

Channel routes and channel/contact contracts.