- Step1: Open Agent.Space and choose an available agent harness, such as Codex or Claude Code.
- Step2: Select a model access plan, using Share for fixed GPT capacity or Flex for metered model usage.
- Step3: Start a cloud workspace and launch the selected agent.
- Step4: Give the agent a task and work with its persistent sessions, files, previews, and team context.
- Step5: Return from another device to continue the work without recreating the environment.
- Step6: Invite collaborators or use the Developer API to connect the plan with another agent product.