Skip to main content
You can talk to Twill from the tools you already use. The important part is that follow-ups stay attached to the same thread (issue/PR/ticket/comment thread), so you can iterate without re-explaining context.

Where you can use Twill

  • GitHub: issues, pull requests, and review threads
  • Linear: issues
  • Notion: comments
  • Slack: mentions (@twill)
  • Sentry: error and issue events
  • AWS: read-only access to cloud resources (logs, infrastructure)
  • Web app: task creation plus logs, previews, and artifacts

What you can do from integrations

Different tools support different triggers, but the workflow is consistent: you create a task, answer questions if needed, and Twill opens a PR for review. You can also use Ask mode to get answers about your codebase without triggering any code changes.
Want a plan first? Use /plan in the message/comment that triggers Twill (or use Linear’s twill-plan label). See Research & Planning.
Use /ask to ask questions about the codebase without making changes. Twill explores the code and answers your question — no PR, no file edits. See Research & Planning.
You can optionally override the agent/model for a new task by adding agent= and/or model= to the message/comment that triggers Twill. This only applies when creating a new task (not follow-ups in an existing thread).

What shows up where

  • Plans and questions are posted back to the thread that triggered the task
  • Implementation happens in a sandbox environment
  • Pull requests are created in GitHub with a summary and proof-of-work when available

How to follow up

Reply in the same thread and mention @twill again. Twill treats it as a continuation of the same task. Examples of clarifying questions:
Twill asking clarifying questions in Slack
Twill asking clarifying questions in the web app