Generative media in your terminal.
A single-binary Go CLI and stdio Model Context Protocol (MCP) server for image, video, audio, and chat generation across Gemini, MiniMax, and Google Flow.
Installs binary to ~/.gflow/bin and adds it to your user PATH automatically.
Supported Backends
Select via --provider or GFLOW_PROVIDER.
| Provider | Modalities | Underlying Models | Transport | Auth Method |
|---|---|---|---|---|
| gemini (default) | Image, Video, Audio, Chat | Imagen 3, Veo (w/ audio & cuts), Lyria, Flash | Direct HTTPS / Headless CDP (port 9223) | Local session token (zero popups) |
| minimax | Video | MiniMax H3 (768P & 2K) | Direct HTTPS | Desktop app token |
| flow | Image, Video, 4K Upsampling | Imagen 4, Veo 3.1 | Direct CDP (port 9222) | Active browser session |
Command Reference
Clean, composable subcommands designed for interactive use and shell scripting.
Generate images via Gemini Imagen 3 (default) or Google Flow Imagen 4.
Generate video clips via Gemini Veo (default), MiniMax H3, or Google Flow.
Generate music and audio tracks using Gemini's Lyria / MusicFX pipeline.
Direct conversational chat in your terminal with Gemini Flash. Prints plaintext to stdout.
Inspect connection readiness across Gemini, MiniMax, and Google Flow.
Launches a dedicated Chrome instance on port 9222 for Google Flow.
Model Context Protocol (MCP)
Connect gflow as a native tool server for Claude Desktop, Cursor, OpenCode, Cline, or Windsurf.
Run gflow mcp setup to automatically configure all installed AI coding assistants (Claude Desktop, OpenCode, Cursor, Claude Code), or run over stdio with gflow mcp.
{
"mcpServers": {
"gflow": {
"command": "gflow",
"args": ["mcp"]
}
}
}