OpenCues patches Gemini CLI so its prompt box gets inline intelligence: cycle word alternatives, fill blanks (5 miles in km _), rewrite your draft (make this more specific _), and run agent tasks, all before the prompt is ever submitted.
Install
Install the CLI with npm install -g opencues, then opencues install gemini-cli. Like every native host, it patches a dedicated copy rather than your main install, and opencues run gemini-cli launches it. Self-healing installs rebuild automatically when the runtime source changes.
Same config as everywhere else
Gemini CLI reads the shared ~/.cues/ directory: the cue you wrote for Claude Code works here unchanged, and a per-host tips pack covers Gemini CLI's own feature vocabulary in context.
Bring any model
Despite the host's name, OpenCues' inline calls use whichever provider you configured: Cerebras, Groq, OpenAI, Anthropic, Gemini, OpenRouter, or local Ollama. The host and the inline assistant are independent choices.
Related: LLM providers · AI in any text field · AI in your terminal · Inline AI in OpenCode · GitHub repo