I use Claude Code and also pay for Perplexity, OpenAI, Gemini, and run Ollama locally. Got tired of switching tabs when the right model for a question wasn't Claude.

Dragoman is a small CLI (~800 lines) that plugs into Claude Code's sub-agent system. Ask about the news → routes to Perplexity. "Try Gemini" → goes to Gemini. "Run locally" → Ollama. You can fan out to four models and have Claude synthesize. Keys resolve from 1Password/Keychain at call time, never enter Claude's context.

The fun part: Anthropic's sub-agent architecture already does the hard work. Dragoman just adds the verb.

Happy to answer questions about the architecture or the sub-agent integration.