Channels

Channels, messages, threads, apps, AI agents, and cross-server sharing

Channels are where conversations happen. But in start.chat, a channel is more than a chat room — it’s a configurable workspace where apps can be pinned, AI agents can participate, data can surface, and workflows can run. The goal is to make each channel feel purpose-built for whatever your team uses it for.

Channel types

Every channel is a combination of a few flags that control its behavior.

Public channels are the default. Any server member can read and write messages.

Private channels restrict access to specific roles. Users without a matching role won’t see the channel in the sidebar or search results. Useful for leadership discussions, HR topics, or paid-tier content in communities.

Read-only channels let everyone view messages but only users with write permission can post. These work well for announcements, changelogs, or status updates where you want visibility without noise.

Solo channels are a unique mode where users can only see their own messages and threads. Think of it like a personal notebook within a shared server — useful for individual task tracking or private AI conversations.

Shared channels bridge conversations between two servers. Messages posted in either server appear in both, making cross-team collaboration possible without requiring everyone to join the same server. See shared channels below.

These flags combine naturally. A channel can be both private and read-only (an announcements channel only certain roles can see), or solo and auto-threaded (a personal journal that organizes each entry into its own thread).

Default channels

When you create a server, several channels are set up automatically:

Server channel (announcements) — A read-only channel that shares the server’s name and ID. Use it for official announcements, updates, or anything you want everyone to see but not respond to directly. Only users with write permission can post here.

#general — An open discussion channel where any member can chat. This is typically the default destination when someone joins the server.

System channels — Behind the scenes, servers also have solo channels for flows, data, and apps. These are hidden from the sidebar and used internally by the platform. You won’t see them unless you’re debugging or building integrations.

Messages

Messages support markdown formatting with code blocks and inline code, file and image attachments with previews, emoji reactions, replies that quote the original message, @user and @role mentions, and automatic link previews that extract titles and images from URLs.

Message creators can edit or delete their own messages at any time. Users with moderation permissions can delete any message. Edits update the message in place with an “edited” indicator.

Message types

Most messages are standard user messages, but the system also supports bot messages (from HUD or external agents, which can stream responses in real time) and draft messages (unsent messages that persist in your editor across sessions).

Keyboard shortcuts

Messages have single-key shortcuts that work when hovering over a message:

KeyAction
PPin or unpin
EEdit
RReact (opens emoji picker)
TReply in thread
DDelete

These complement the message menu, which also includes reply, copy link, copy text, and mark read/unread.

Threads

Threads branch off from any message to keep focused discussions organized without cluttering the main channel. Click a message to start a thread — the original message becomes the thread root, and replies appear in a side panel.

Threads get AI-generated titles based on the conversation content, making them easy to scan and find later. Titles update as the discussion evolves. Reply counts, update timestamps, and unread indicators show which threads are active at a glance.

Auto-threading

Any channel can be set to auto-thread mode, which automatically creates a thread for every new message. This transforms a channel into a forum-style view where each post gets its own discussion — useful for support channels, feature requests, Q&A, or any context where you want organized, topic-per-thread conversations instead of a single stream.

Pins

Important messages can be pinned to a channel. Pinned messages are accessible from the channel header, giving your team a quick-reference list for important decisions, links, or announcements. Pin and unpin with the P shortcut or from the message menu.

Full-text search works across all messages and data you have access to. Results are ranked by relevance and can be filtered by channel, user, and date range. Search indexes messages as they’re sent, so new content is immediately findable.

AI agents in channels

Channels can have one or more AI agents configured with independent response modes. The built-in HUD agent is available everywhere by default, and you can add external agents (like OpenClaw) that connect via webhook or websocket.

Each agent’s response mode controls when it participates:

ModeBehavior
AllResponds to every message inline
All threadsAuto-replies in threads
Threads onlyOnly responds inside threads
Mentions onlyOnly responds when @mentioned
OffDisabled for this channel

Having multiple agents in the same channel lets you compose different capabilities — one agent for code review, another for customer lookup, a third for task management — all responding in the same conversation.

Per-channel AI configuration also supports custom system prompts, model selection, and ability toggles, so the same HUD agent can behave differently in your #support channel versus your #engineering channel.

Channel settings

Channel settings are accessible from the channel header. Available options depend on your role permissions.

Name, description, and slug. The slug determines the #channel mention format and URL path.

Topic. A short description displayed at the top of the channel. Channels can have multiple topics, and messages can be tagged with topics to organize conversations by subject.

Slowmode. Rate-limits how often members can post (5 seconds to 10 minutes). Useful for high-traffic channels where you want to encourage thoughtful messages.

Max message length. Caps messages at 500, 1000, 2000, 4000, or 8000 characters. Defaults to 4000.

Agents tab. Configure which AI agents respond in this channel and how. See AI agents in channels above.

Permissions tab (private channels). Grant or revoke access by role.

Notifications

Each user can independently configure notification preferences per channel:

SettingBehavior
All messagesNotify on every message
Mentions onlyNotify only when @mentioned
NoneSilent — no notifications

Users can also temporarily mute a channel until a specific time. Notification preferences are per-user and don’t affect other members.

Channel groups

Channels in the sidebar can be organized into collapsible groups (like “Engineering”, “Marketing”, “Off-Topic”). Server admins can create, rename, and reorder groups, and drag channels between them. Ungrouped channels appear at the top level.

Shared channels

Shared channels let two servers communicate without merging. An admin in one server initiates a shared channel, and admins in the target server receive a notification to accept or decline. Once accepted, messages are mirrored bidirectionally — each server maintains its own copy of the channel with independent permissions and notification settings.

This is useful for partnerships, open source project collaboration, customer support bridges, or any situation where two communities want a direct line without giving up their own server structure.

Indicators

Typing indicators show when other users are composing a message in the current channel or thread. Read status tracks which messages each user has seen, powering unread counts and the “new messages” divider.

Programmatic access

The Channel API covers channel management. Messages, Threads, Reactions, Attachments, Pins, Channel Topics, and Notification Settings each have dedicated endpoints. All are available via the REST API, SDK, and CLI.