Gemini CLI silent at startup? Find which stage is waiting.

A live process with no output does not identify the cause. Determine whether the CLI reached its prompt or sent a request, and preserve configuration before changing it.

Do this first: Run a minimal non-interactive startup, inspect local config and shell environment, and reset only the affected configuration. Verify the project diff after any restart.

What the failure looks like.

Gemini CLI launches, remains alive, and produces no usable prompt or error. The terminal looks frozen even though the process has not exited.

Why it happens.

Corrupted local configuration, shell initialization, or a startup dependency can block the CLI before it reaches a model request.

Fix it without losing the task.

Determine whether the process is alive, whether a prompt appeared, and whether any request was sent. Record the exact command and elapsed time.

Check the active configuration path, shell startup output, authentication state, and network errors one at a time. Do not reset all settings before preserving the current configuration.

Use the linked issue only as a comparison if the symptom and affected version match. A similar silent startup does not prove the same cause.

After a controlled restart, confirm the CLI reaches an input prompt before resuming the coding task. Then inspect the project diff separately.

What the next AI should receive.

Exact startup command and shell, CLI version, active config path, last startup output, whether a prompt appeared or a model request was sent, and any preserved configuration change.

Check before restarting

Check whether Gemini CLI has reached its prompt or issued a model request. Capture the startup command and relevant local logs before changing configuration.

Boundary of the evidence

Repository state cannot identify a CLI startup root cause by itself. The issue cited here is an individual report, not proof that the behavior is widespread.

Diagnostic checkpoint

The key diagnostic split is whether the CLI reaches its prompt or sends a request; preserve config before testing either path.

How ShardStitch helps.

ShardStitch can preserve the repository task while you diagnose Gemini CLI startup separately. Include the command and configuration details in the handoff; the packet does not read the CLI logs unless you provide them.

Source trail.

Gemini CLI issue #12163, the issue tracker, and the Gemini CLI repository.

FAQ.

How can I tell a Gemini CLI startup hang from a slow model response?

Record whether the CLI reached its prompt and whether a request was sent. A failure before the prompt points to local startup; a request already in flight needs request/log evidence.

Should I delete Gemini CLI configuration to make it start?

Not as a first step. Preserve the active configuration and inspect the exact startup error. Change or reset only the setting supported by the diagnostic evidence.

Scope note: Evidence boundary: the linked Gemini CLI issue is one report. Treat shell, configuration, authentication, and network explanations as hypotheses until local logs identify the failing stage.