Best for workspace-aware edits plus an integrated terminal test loop.
Cursor local GCP sandbox
Cursor is useful when the agent can reason over an IDE workspace and edit files quickly. LocalCloud gives those edits a concrete local GCP target so agent-written code can be exercised through Google Cloud SDKs before a human pushes to a real project.
Keep emulator env vars in the terminal/session running tests.
Use repository-local instructions to prevent real-cloud fallbacks.
The editor and terminal may not share state
Cursor can edit files and run commands, but environment variables must be present in the process that executes tests. Have the agent echo emulator env vars immediately before the SDK check.
Prefer a project-local note once proven
After the first successful LocalCloud run, ask Cursor to add concise project instructions only if the repository already uses an agent instruction file. The instruction should say localhost endpoints are the default local test target and production deploys require real GCP validation.
Commands and prompts
Use these snippets as starting points, then keep the checks scoped to localhost until you intentionally validate against real Google Cloud.
Inspect this workspace for Google Cloud SDK usage. Use https://local.cloud/ai/agents.md to start LocalCloud, set emulator env vars in the terminal that will run tests, and validate one local operation. Do not add real GCP credentials or production project IDs. Use the sandbox deliberately
- Cursor-specific background agent environments can differ from the visible terminal.
- Do not store fake local env vars in files that production jobs source without safeguards.
Keep the agent on the supported path
- Compatibility matrix — Check current support boundaries before relying on a local-only test.
- Service catalog — Review every LocalCloud service, endpoint, and limitation.
- SDK examples — Use standard Google Cloud SDKs pointed at localhost.
- Seed data — Load deterministic fixtures for repeatable agent and CI runs.