Skip to main content
Twill runs the coding agent you already use — Claude Code, Codex, or OpenCode — wherever you are. Start a task on your Mac, hand it to the cloud mid-task, follow it from your phone, or kick it off from a Slack thread or a GitHub issue. Every conversation and result lands in the same place.

Local and cloud

A local project is a set of folders on one connected computer. A cloud workspace holds a team’s repositories, environment, integrations, billing, and tasks; new cloud tasks normally run in a fork of its main environment. A task is a chat. Each message starts a run of the agent. Follow-ups keep the task’s harness and checkout; you can change models within that harness. Locally, the harness uses your native configuration; in the cloud, Twill configures it with workspace credentials, integration tools, and verification instructions, so available commands and models can differ. Handoff has the local agent push its work and brief a linked cloud task. Continue locally brings pushed branches and PRs back. Live processes, local databases, and native logins don’t travel either way.
Conversation history and agent output sync to Twill; local execution is not an offline mode.