What to run on a Pod.
Run Claude Code and Codex on Pods for parallel branches and Simulator tests. Use a Claude Code workflow for PRs from Slack, or separate environments for client projects.
Each case below says what it needs, what it costs and where to start.
Parallel feature branches
One developer, several agents, one Pod. Use separate worktrees for a refactor, a new screen and test fixes. Two xcodebuild runs execute at once; further builds queue. Claude Code and Codex sessions can share the same Pod; Limits and queues has the numbers.
- Needs: a connected Claude or Codex environment; start from either quickstart below.
- Capacity: two to three active agents per standard Pod, two builds at a time.
- Costs: Pod runtime, metered per second, whatever the agent count.
- Result: one branch and one PR per agent, commits attributed to you.
UI test sweeps on Simulators
Have an agent run a UI test flow on iPhone and iPad Simulators, compare screenshots with the previous run and report what changed. Each Pod has three to four pre-booted Simulators, leased per agent for a run.
- Needs: a UI test target or a scripted flow the agent can drive through XcodeBuildMCP.
- Capacity: three to four Simulators per Pod; current iOS and the previous major on the Latest Xcode.
- Costs: Pod runtime, metered per second, until the Pod stops ten minutes after its last session.
- Result: screenshots and logs in the dashboard and attached to the PR.
PRs from Slack, with Claude Code
An example: someone in #ios-releases tags Claude with a crash. The agent gets a Pod, reproduces the crash in a Simulator, fixes it, runs the UI tests and opens a PR with screenshots attached. The developer reviews the PR and the screenshots.
- Needs: a Claude Team or Enterprise organization with a self-hosted environment connected to Macswarm, and Claude in your Slack.
- Capacity: one Pod per developer who triggers work; the plan's Pods-at-a-time limit applies.
- Costs: Pod runtime, metered per second.
- Result: a PR with screenshots, commits attributed to the developer who asked.
Separate environments for each client
Keep one environment per client, with its own Pods and keys. Pods are cloned from the golden image and destroyed ten minutes after their last session ends; logs and screenshots stay with the environment they came from. The dashboard reports usage by environment, so you can allocate Pod-hours to each client.
- Needs: one connected environment per client.
- Capacity: Team allows ten Pods at a time across environments; Enterprise reserves Macs for you.
- Costs: Pod runtime per environment, shown in the dashboard.
- Result: client work that never shares a Pod, a key or a log with another client.
Run your first task on a Pod.
3 Pod-hours free, no card. Connect Claude Code or Codex and the first Pod boots from the warm pool.