@tailored-ai/channel-slack

Slack channel for Tailored AI. Ships as a register(ctx) plugin (#47) — the host invokes the package's default export with a PluginContext during runtime construction; the plugin registers a channel factory the runtime starts automatically when channels.slack.enabled: true is set in config.yaml.

Designed as a reference implementation of the channel registry (#81) — small enough to read in one sitting (~250 lines) but functional enough to be useful day-to-day. The deeper Discord channel in @tailored-ai/core is the feature-complete version; treat this as the canonical example for anyone authoring a third-party channel.

Install

tai plugin install @tailored-ai/channel-slack

Then in config.yaml:

plugins:
  - "@tailored-ai/channel-slack"

channels:
  slack:
    enabled: true
    token: ${SLACK_BOT_TOKEN}      # xoxb-... bot user OAuth token
    appToken: ${SLACK_APP_TOKEN}   # xapp-... app-level token with connections:write
    respondToDMs: true
    respondToMentions: true
    # Optional — restrict to specific workspaces
    # allowedTeams:
    #   - T01234567
    # Optional — bind specific channels / DMs to projects
    # projectMappings:
    #   - channel: C01234567
    #     project: my-project
    #   - dm: true
    #     project: personal-notes

Slack app setup

This MVP uses Socket Mode, which means no public webhook is required.

1. Create a Slack app at https://api.slack.com/apps.
2. Under Socket Mode, enable it and generate an app-level token with connections:write — this is your SLACK_APP_TOKEN.
3. Under OAuth & Permissions, install the app to a workspace. Add these bot scopes:
   • app_mentions:read
   • channels:history, groups:history, im:history, mpim:history
   • chat:write
   • im:write
   • users:read
4. Copy the Bot User OAuth Token — this is your SLACK_BOT_TOKEN.
5. Under Event Subscriptions, enable events and subscribe to bot events: message.channels, message.groups, message.im, message.mpim, app_mention.

Behavior

• Direct messages route to the agent loop directly. The bot's reply is posted in-channel (the DM).
• Channel messages trigger a response only when the bot is @-mentioned. The reply is threaded under the original message.
• The bot mention is stripped from the prompt before the agent sees it.
• Concurrent messages from the same user (per project) are queued — the bot replies "I'm still working on your previous message" until the first finishes.
• Long responses (> 3000 chars) are split on newline / space / hard boundaries.

What this MVP skips

Deliberately out of scope so the file stays readable:

• Slash commands (/new, /agent, /context, /tasks, config-driven commands) — see Discord's syncCommands for the reference shape.
• The built-in /context + /tasks handlers — those would copy ~200 lines from Discord without learning value.
• Multi-user per-agent state — the agent selection is global / config-driven.

Contributions for any of these are welcome.

License

MIT