Set Up Gemini CLI on a Shared Research Workstation

A shared research workstation gives a team convenient access to datasets, analysis tools and compute. Adding Gemini CLI can help researchers explore unfamiliar code, explain errors and draft analysis scripts. A reliable setup starts with separate user accounts, a verified API connection and clear boundaries around the files each session can access.

01 · 1. Give each researcher a separate workstation login

Use an individual operating system account for each researcher. Separate project folders alone do not isolate credentials when everyone uses the same login. Keep personal shell configuration, Gemini CLI settings and API keys inside each person’s home directory.

Before installing anything, agree on which research files may be sent to an external API. Start with a small working directory containing approved material. Keep participant records, private credentials and unrelated datasets outside that directory.

02 · 2. Check Node.js and install Gemini CLI

Run node --version and npm --version in a terminal. Check the current Gemini CLI installation requirements and use a supported Node.js release. On a centrally managed workstation, ask the administrator to provide the runtime if it is missing.

For a personal installation on a Linux workstation, create a user-owned package location with mkdir -p "$HOME/.local", then run npm install --global --prefix "$HOME/.local" @google/gemini-cli. Add that directory to the current shell with export PATH="$HOME/.local/bin:$PATH".

Run gemini --version to confirm the executable works. Add the PATH setting to your own shell startup file if needed. Record the installed version in the team’s setup notes so colleagues can reproduce the environment.

03 · 3. Prepare Quanta credentials and verify the endpoint

Quanta fits teams running sustained workloads across several tools: one pay-as-you-go balance covers supported models and works with Gemini CLI, Claude Code, Codex and OpenAI SDK clients. Open a Quanta account, then obtain an API key through its account interface.

Quanta’s API base address is https://api.quantarelay.com. Before configuring Gemini CLI, check Quanta’s current Gemini CLI instructions for the exact Gemini-compatible endpoint and authentication settings. Protocol compatibility matters: Gemini CLI’s native API connection needs the Gemini-compatible route, and an OpenAI-compatible route alone is insufficient.

Use a separate key for each researcher when the account supports it. Otherwise, document who owns the shared credential and how the team will replace it when access changes.

04 · 4. Load the key into your own terminal session

In Bash, enter read -r -s -p 'Quanta API key: ' GEMINI_API_KEY, press Enter after pasting the key, then run export GEMINI_API_KEY. This avoids placing the key itself in a shell command that could enter command history. The hidden input prompt will not display the pasted value.

Set GOOGLE_GEMINI_BASE_URL to the exact endpoint specified by Quanta’s Gemini CLI instructions. Do not guess a version suffix. Launch gemini and choose API-key authentication if prompted. Avoid placing credentials in project repositories or shared shell startup files.

05 · 5. Test one small request before opening research files

Start Gemini CLI from an empty test directory. Select a Gemini model available through your Quanta account, such as Gemini 2.5 Flash, using the exact model identifier Quanta documents. Ask for a short explanation of a basic programming concept.

Check that the response arrives and that the request appears in Quanta’s usage records. If authentication fails, check the key, endpoint and selected authentication method. If model selection fails, verify the identifier and account availability before changing other settings.

06 · 6. Establish file access and spending habits

Move into the approved project directory only after the connection works. Review proposed shell commands and file changes before approving them. Begin with a task such as explaining one script, then inspect the result before requesting broader changes.

• › Keep automatic approval disabled while establishing the workstation workflow.
• › Review token usage after representative tasks; Quanta meters real tokens and prices cache writes and cache reads separately.
• › Record working CLI versions, endpoint settings and model identifiers without recording secrets.
• › Run unset GEMINI_API_KEY when finished and close the session. Use the team’s approved secret storage for future sessions.