Overview
Run coding agents in cloud sandboxes from the dashboard or API.
Ellipsis runs Claude Code and Codex in cloud sandboxes with your repositories, dependencies, and credentials. Start work from the dashboard or API, or save an automation that runs on a schedule or event.
| Resource | Use it for |
|---|---|
| Session | One task or conversation with an agent |
| Environment | Repositories, dependencies, secrets, and compute shared by sessions |
| Automation | Saved instructions and session settings, with an optional trigger |
| Code Review | Reviewing new commits on pull requests |
Start a session
Choose a repository and ask the agent to make a change. Follow its work, inspect the diff, and continue the conversation from the dashboard.
Configure reusable work
Commit YAML under .ellipsis/ on a repository's default branch. Environments define where work runs; automations define the work.
ellipsis:
name: check-tests
session:
harness:
type: claude_code
instructions: |
Run the test suite in api-repo. Report failing tests and
their errors. Do not modify code.
environment:
repositories:
- name: api-repo
permissions:
github:
permissions: read_only
budget:
session: 3Invoke check-tests from Automations in the dashboard or through the API. Each invocation starts a session with these settings.
Build with the API
Use the REST API, Python SDK, or TypeScript SDK to start sessions, send messages, stream records, and run automations.
Check model support before selecting a harness or adding capabilities.