A computer of its own, prepared for the job.
Every AI employee gets its own computer, prepared for its role, rather than one machine shared by all of them. Give each one your code, the software its job needs, and its own accounts, with access you control.
See how it’s prepared- GitHub
- Slack
- Your repositories
- App servers
- Passwords & secrets
- yourcompany/your-repo-1
- yourcompany/your-repo-2
- Software and services for your project
Prepare the computer once. Reuse the setup.
Set up the project’s computer with the tools it needs, then save that setup. Later tasks can restore it instead of repeating the heavy preparation.
Prepare the project
Choose the software, check out the repositories, and configure the services the job needs.
Save the starting point
Capture the prepared setup as a snapshot that can be restored for later work.
Start the next task
Restore the saved setup, start the configured services, and bring the app to readiness.
See example environment configuration
For the technical setup, a manifest describes the operating system, repositories, installation commands, and services. These Linux and macOS examples show how the pieces fit together.
agent-env.yml
version: 1
platform:
os: darwin/arm64
sources:
- repo: acme-corp/mobile-app
workspace: /Users/admin/workspace/mobile-app
setup:
- name: Install dependencies
run: npm install
services:
- name: api
run: npm run serve
ports:
- port: 8080
name: http
healthcheck:
url: http://localhost:8080/health
interval: 2s
timeout: 30s
simulators:
- platform: ios
agent:
instructions:
- .claude/CLAUDE.mdRunning in Wallfacer VM
Give it the tools to check its work.
A coding AI employee needs to run the project, see the result, and investigate what went wrong. Its own computer can hold the same tools your team uses. Another role needs another set of tools, so it gets its own computer, prepared for that job.
Web apps and services
Prepare Linux with your app, backend, database, and browser tools. The AI employee can start services, run tests, and click through the changed flow.
Native and iOS apps
Prepare macOS with Xcode and the iOS Simulator. Give that AI employee the tools to build your app and inspect it on a simulated device.
Your project, together
Keep related repositories, project dependencies, and configured MCP tools on that role’s computer. Choose what the job needs during setup.
Before requesting review
Run the checks
Build the project and run the tests in your handbook.
Look at the result
Open the changed flow in a browser or Simulator.
Bring back the evidence
Attach findings and screenshots. Explain anything that could not be verified.
Give access to the AI employee doing the work.
Choose which repositories and services each AI employee can use. Its working identity supplies the access for the session.
Explore controls and work recordsCredentials belong to the session.
Per-identity credentials are supplied for the work and scrubbed before a reusable snapshot is captured.
Reasoning and tools run separately.
The agent uses tools on the prepared computer through a controlled connection. Your code and credentials live on that computer, separate from the reasoning process.
Need the computer on your own infrastructure?
Talk with us about enterprise deployment on your hardware or cloud, including private networking and dedicated macOS hosts. We’ll work through where the code runs and what your team needs to manage.
Let’s prepare your AI employee’s workspace.
Tell us what the job needs to run and which tools your team uses. We’ll work through the computer, access, and setup together.
Talk about your setup