Skip to main content
Once your workspace is set up, most day-to-day use happens directly in Slack. You can start a conversation with an @coderabbit mention, through Slack assistant entry points (when enabled), or with slash commands like /plan, /learn, and /automations.

​
What each command is for

Entry pointWhat it doesDetails
@coderabbit mentionInvestigate, explain, plan, or act on anything in the current conversation. The run uses the scope matching the current channel, thread, or DM.Scopes
/planGenerate a structured implementation plan from a request or thread. If used outside a thread, Slack Agent creates one automatically.Plans
/learnCapture a durable fact from the current thread or from text you provide directly.Knowledge Base
/automationsList, create, or manage recurring tasks for the current channel or DM.Automations

​
How conversation behavior works

Slack Agent behaves differently depending on the Slack surface:
Each top-level DM starts a new conversation. Threaded replies continue that same conversation. In single-player threads (just you and Slack Agent), there is no need to re-mention the bot. It stays active automatically.
Slack Agent uses the scope matching the current surface. You never need to select a scope manually. The matching scope determines repositories, connections, spend limits, and knowledge privacy automatically.
If multiple scopes match the current surface, requests are blocked as ambiguous rather than auto-selecting one. Learn more about scope resolution and ambiguous matches.

​
What determines what Slack Agent can use

Each run is governed by the scope that matches the current Slack surface. That scope determines:
  • Which repositories are available
  • Which connections can be used
  • What spend limit applies
  • Whether the shared sandbox is available
  • Whether knowledge should be shared or private

​
Who can use Slack Agent

Slack Agent responds to workspace members only.
External users (guests) are blocked by default. Channels marked as externally shared are fully unsupported, and all bot usage is blocked in those channels.

​
After a run

After completing a task, Slack Agent replies in the thread with its findings, a proposed plan, follow-up actions like opening a PR, or an offer to save what it learned to the Knowledge Base. For deeper inspection, the companion web app provides thread reviews for run-by-run execution detail, usage for workspace-wide activity, and plans for version comparison and agent handoff.

​
Slack vs. the web app

Use Slack when you want to…Use the web app when you want to…
Ask Slack Agent to investigate, plan, learn, or act in contextInspect what happened after a run in more detail
Manage automations from the conversation where they applyConfigure scopes, connections, or sandbox settings
Confirm knowledge proposals from the same threadBrowse knowledge history, plans, usage, and thread review artifacts

​
What’s next

Scopes

Learn how Slack Agent decides which repositories, tools, and limits apply in each conversation.

Automations

See how recurring Slack Agent tasks work for channels and other Slack surfaces.

Plans

Review how /plan work moves from Slack into a structured implementation plan and handoff flow.