Skip to main content
Capy cloud agents are in alpha. To add them to your organization, email [email protected].
Revyl can hand work on a screen to a Capy agent. Pick a screen in your app’s Atlas, describe what you need, and Revyl starts a thread in one of your Capy projects with that screen as context. The agent can run your app on a Revyl device, and what it starts there is attributed back to the conversation in Revyl.

​
What it does

  • Send an Atlas screen to Capy. Start a Capy agent from a screen’s focus card, or by mentioning @Capy in a comment on the screen. You describe the task, and Revyl starts the Capy thread.
  • Track the conversation from Revyl. Every launch shows up under Cloud Agents in the sidebar, next to Cursor conversations, with a detail page and Open in Capy.
  • Hear back where you asked. An agent started from a comment posts one final update to that comment thread.

​
Before you start

  • Capy cloud agents are turned on for your organization.
  • An owner or admin has connected Capy. Launches from Atlas use the organization’s Capy connection.
  • At least one enabled Capy project checks out the app’s repository. When Revyl knows the app’s repository, it preselects the project that checks it out, and asks you to enable one if none does.

​
Launch a Capy agent

​
From a screen

Open your app’s Atlas and select a screen. In its focus card, choose Capy agent. In Spawn a Capy agent:
  1. Describe the task. You can also mark regions on the screenshot.
  2. Check the Project, Repository, and Model, and the device and any launch configurations.
  3. Click Send to Capy.
While the agent runs, the panel shows its status, Open in Capy, and Stop. The focus card lists every agent started from that screen under Agents spawned here.

​
From a comment

In a screen’s comment thread, mention @Capy and describe the task. Before the first send, Revyl shows the same project, repository, model, and device settings. Capy replies in the thread when it’s done. Mention @Capy again in that thread to follow up in the same Capy thread.

​
What the agent gets

Revyl starts the Capy thread with a prompt built from:
  • your request
  • the screen you picked and its Atlas evidence
  • how to install and use the Revyl CLI
  • the device to use: the one you chose, or Auto
  • the names of any launch configurations you selected; their values stay in Revyl
  • from a comment: the thread to reply in
The prompt carries no secrets. Capy’s thread environment carries a Revyl credential issued for you, so the agent can use the Revyl CLI, and Revyl attributes the device sessions, builds, tests, and workflows it starts to the conversation. Those resources link back with Open in Capy.

​
How it differs from Cursor cloud agents

  • Starts from Atlas only. Capy agents launch from Atlas screens and their comments.
  • No starting branch or automatic pull request. Capy’s launch has no Start from or pull request option.
  • Messages stay in Capy. Revyl shows the conversation’s status and resources, and links to the Capy thread for its messages.
  • One connection. Launches use your organization’s Capy connection. Cursor has a personal key and a team default.