imagine-mcp 1.0.0

MCP server for image/video understanding & generation (Gemini/OpenAI/Grok)

imagine-mcp

Production-grade MCP server for image/video understanding and generation across Gemini, OpenAI, and Grok.

Features

• 4 MCP tools: understand, generate, config, help
• 3 providers: Google Gemini, OpenAI, xAI Grok — same interface, switch via param
• 2 tiers: poor (cheap/fast) and rich (high quality) per provider
• Image + video: understanding (Gemini native multimodal) and generation
• Zero-config onboarding: browser-based credential setup via relay (no .env)
• Degraded mode: server starts with 0 credentials; configure via config tool
• Cache: automatic caching of understand responses (configurable TTL)
• Leaderboard-ranked models: provider ordering auto-refreshed weekly from Artificial Analysis + LMArena

Quick start

# With uvx
uvx imagine-mcp

# Or via MCP client config (Claude Desktop / Cursor / etc.)
{
  "mcpServers": {
    "imagine": {
      "command": "uvx",
      "args": ["imagine-mcp"]
    }
  }
}

On first run, open the relay URL from stderr and submit your API keys.

See docs/setup-with-agent.md and docs/setup-manual.md for onboarding details.

Model IDs

See docs/models.md (auto-generated, leaderboard-sorted) and CLAUDE.md for the full provider × action × tier → model ID mapping.

Development

git clone https://github.com/n24q02m/imagine-mcp
cd imagine-mcp
mise run setup    # install tools, deps, pre-commit hooks
mise run test     # run test suite
mise run dev      # run server in dev mode

License

MIT. See LICENSE.