podbaydocs

What Claude can do in a pod

See the useful outcomes Claude can deliver when it has a persistent computer, project tools, and Podbay's built-in capabilities.

A pod gives Claude more than a place to edit files. It gives Claude a computer where it can turn a request into a working, inspectable system—and return to that system later.

You do not need to learn Podbay's internal commands. Describe the outcome, constraints, and approval boundaries. Claude uses the tools available inside the pod to work out the implementation.

Build and run the whole project

Claude can work across the parts that usually surround an application, not only its source code:

  • install project packages and use the environment's language and framework tools;
  • run the application with Postgres, workers, queues, or other supporting services;
  • apply database migrations, seed development data, and verify that the pieces work together;
  • keep important services running after a crash or pod restart;
  • give you a live, owner-only link to inspect the result.

Get this project running with its database and worker. Fix the setup problems, verify the main flow, and send me a private link when it is ready.

Receive work in a form you can open

The pod is a remote computer, so its local file paths are not useful on your phone or laptop. Ask Claude to return the result through the right channel:

  • a live preview for an application or dashboard;
  • a GitHub link or pushed branch for committed source changes;
  • a summary, table, or short document directly in the Claude conversation;
  • an exported or backed-up copy when the result is data that must outlive the pod.

Put the finished guide in Git, send me a link I can open, and summarize the decisions here.

Continue the same work from Claude

The project, installed tools, working notes, and Claude session stay with the pod. Start from the official Claude desktop app, close your laptop, and continue from Claude on mobile without moving the workspace onto either device.

Give Claude work that repeats

Claude can perform recurring work with the same project context and tools it uses during an interactive session. Useful examples include:

  • preparing a morning research brief;
  • checking a set of sources and reporting meaningful changes;
  • refreshing data and updating a private dashboard;
  • running a weekly project review or maintenance routine;
  • monitoring a workflow and surfacing when it stops reporting.

Podbay keeps a run history and can surface stalled work. The pod must remain running for scheduled work to happen on time.

Research beyond the pod

Claude can read most public websites directly. When a permitted site rejects cloud datacenter traffic, the optional Podbay Relay can let Claude make the request through your own computer and network. You can see how the relay is used, pause a pod, block a site, and decide which sites may use your signed-in browser session.

Ask another pod for help

If you keep different projects in separate pods, Claude in one pod can ask Claude in another to do project-local work and report back. This is useful when one result depends on another workspace—for example, asking a website pod to publish new documentation while a research pod prepares the source material.

Messages stay within your own Podbay account. The receiving agent still follows its normal approval rules; a message is context, not permission to publish, spend money, or change an external system.

Start with prepared expertise

An environment can give Claude the right toolchain, project rules, templates, and job-specific skills before the first conversation. Some environments also include opinionated workflows such as source monitoring or framework-specific testing. You can inspect and change the workspace after launch.

Recover common failures

Podbay watches the Claude session, disk, scheduled work, the development server, and services Claude has marked as important. It can restart several common failures and give Claude a diagnosis when something needs attention.

What you still control

Claude has broad freedom inside its pod. You remain the decision-maker for the outside world:

  • only you add or change project secrets;
  • previews stay private until you make them public;
  • pushing code, publishing, spending money, or contacting people still requires the relevant permission;
  • deleting a pod and its stored workspace is an owner action.

The useful way to ask

State the result, the important constraints, and what needs your approval. You do not need to tell Claude which Podbay command to run. For recurring work, also include the timing, timezone, sources, and what should count as worth reporting.

Go deeper