Sazabi
Sandbox CLIs

Gemini

Give the Sazabi agent the Gemini CLI (gemini) in its sandbox for AI-powered coding with Google's Gemini models.

About

This CLI connection makes the Gemini CLI (gemini) available in the Sazabi agent's sandbox. The agent can use it for AI-powered coding. The tool is baked into the sandbox image; connecting it authenticates the already-installed CLI.

Prerequisites

  • A Sazabi project.
  • A Gemini API key (GEMINI_API_KEY), available from Google AI Studio.

Set up in the dashboard

Configure the sandbox in the dashboard under Settings > Sandbox CLIs.

Find Gemini in the catalog

In Settings > Sandbox CLIs, find Gemini under Browse sandbox CLIs and choose to connect it.

Provide credentials

Enter the required value: GEMINI_API_KEY — your Gemini API key from Google AI Studio. Save to store it securely; Sazabi injects it into the sandbox when the agent runs.

Set up with the CLI

You can also configure the Gemini CLI connection with the Sazabi CLI (installed and authenticated — see CLI reference).

sazabi sandbox-clis set --type gemini-cli --env GEMINI_API_KEY=<your-gemini-key>

To validate the credentials without saving, run the same command with test in place of set (the test command also requires the --env flags).

Verify

Ask the agent, in a thread, to run the gemini CLI (a read-only or status command) and confirm it returns real output from your account. If the command is not found, the CLI connection may not be enabled; if it runs unauthenticated, recheck the credentials.

Troubleshooting

Command not found — Confirm the Gemini CLI connection is enabled for the project under Settings > Sandbox CLIs.

Authentication fails — Recheck the credential values; a rotated or revoked token must be re-entered under Settings > Sandbox CLIs or with sazabi sandbox-clis set --type gemini-cli.

Further reading