Gemini CLI is not acting like a side project. ToolVitals puts it at 97 overall, 95 health, and a perfect 100 shipping score. The repo has 102,377 stars, 30 releases in 90 days, and 34 release events in the last 30 days. That is not casual maintenance. That is a team pushing hard on the product surface.

The official positioning is blunt. The site says Gemini CLI helps you build, debug, and deploy with AI from the terminal, query and edit large codebases, generate apps from images or PDFs, and automate complex workflows with Gemini 3. The GitHub README tightens the pitch further. It calls the project an open-source AI agent for the terminal, and it claims a free tier, 60 requests per minute, 1,000 requests per day, a 1M token context window, Google Search grounding, shell commands, file operations, web fetching, and MCP support.

The interesting signal is not just that it ships. It is what the releases focus on. v0.39.0 added a /memory inbox command, a memory usage test harness, and sandbox cleanup work. v0.40.0-preview.2 added retries for OpenSSL 3.x streaming errors, prevented YOLO mode from being downgraded, bundled ripgrep for offline support, and made base URL handling respect environment settings. v0.40.0-preview.4 was a follow-on patch release. This looks like a team turning an agent into an actual daily tool, not a demo.

What the data does not tell you is just as important. ToolVitals sees commits, releases, stars, SSL, and uptime. It does not see code quality, user satisfaction, revenue, or whether the product actually works well in a real team’s workflow. A fast release train can hide messy internals just as easily as it can signal momentum.

Compared with adjacent tools, Gemini CLI is unusually busy. LangChain shows 32 release events in 30 days and 134,895 stars, so Gemini CLI is slightly faster on release cadence but smaller in audience. OpenClaw is much larger at 363,871 stars, but it logged 28 release events in 30 days, which makes Gemini CLI the more aggressive shipper in this group. The pattern is clear, even if the market outcome is not.

If your team wants a terminal-native AI agent and you care about memory, offline support, and rapid iteration, Gemini CLI is worth a trial now. If you need proof that the product is broadly loved or operationally boring, keep watching. The shipping is real, but the usage story still needs evidence.

Sources:

Sources