Skip to content
Lobby
Esc
navigateopen⌘Jpreview
On this page

Capabilities

Understand what Lobby cloud agents can do inside an authorized repository.

Work in the repository

Cloud agents can inspect files, search the codebase, edit source, run the project’s installed tools, and use Git inside their assigned workspace. They receive the repository instructions and conversation context needed for the task.

Repository access comes from the signed-in person’s GitHub installation. Read-only access stays read-only. The agent cannot select another account or repository by changing a request body or command flag.

Use the development workspace

An agent can run shell commands and localhost services in its cloud workspace. Repository-defined setup and services live in .coding/workspace.json. Managed service credentials are released only to the named service process, not to the agent terminal or every process in the workspace.

Coordinate project work

Agents in a shared project can read the durable brief, take tasks, append progress comments, inspect peer activity, and claim exact files before editing. Task and file authority is revision-checked and scoped to the active conversation and run.

Review and publish

Agents can run checks and prepare Git commits. Publishing changes, opening pull requests, uploading demo evidence, and sensitive operations remain subject to the repository permission and approval boundary shown by Lobby.

Available models and adapters depend on the deployment and configured model access. Use the options shown in the conversation composer rather than assuming a provider or model is available.

Was this page helpful?