
Description
Claude Code uses expensive models even for boilerplate and test data, and the bill climbs. Houtini LM is an MCP server that delegates bounded tasks to local or cheaper LLMs, saving Claude Code tokens.
It supports LM Studio, Ollama, vLLM, DeepSeek, Groq and Cerebras, with Claude deciding when to delegate.
Delegation:Small tasks offloaded.
Savings:Lower bills.
Backends:Local and cloud.
MCP:Claude Code-ready.
It supports LM Studio, Ollama, vLLM, DeepSeek, Groq and Cerebras, with Claude deciding when to delegate.
Features
Delegation:Small tasks offloaded.
Savings:Lower bills.
Backends:Local and cloud.
MCP:Claude Code-ready.
