# claude-code-delegate-local MCP server

> Delegates Claude Code subagents to local models, DeepSeek, AWS Bedrock, or OpenAI-compatible backends without losing the orchestrator session.

- Repository: https://github.com/fegone/claude-code-delegate-local
- Page: https://hostedmcp.org/servers/fegone/claude-code-delegate-local
- Transport: —
- Stars: 19
- License: MIT

## Tools (2)

- `delegate_to_gemini`
- `submit_dsh`

## FAQ

### Can I use local models to save tokens on my Anthropic Max plan?

Yes, you can route specific subagents to local models like Ollama or llama.cpp to save tokens while keeping your main Claude Code session on your Max plan.

### Does this server support DeepSeek's thinking mode?

Yes, it supports dual-format backend routing which works with DeepSeek's reasoning_content thinking mode out of the box.

### How does the server handle full tool calling for delegated agents?

Delegated agents receive the same tool calling capabilities as native Claude Code subagents, including read_file, write_file, and run_bash.
