Not written by an AI, so bear with me :P
Has anyone else tried to use their local LLM in conjunction with Claude Code?
I looked into Pi.dev a bit, and from their documentation, I read about their RPC Mode which allows me to send command line commands to it.
So I'm thinking of making a MCP Claude Code can utilize to use Pi.Dev as a sub-agent and save a bit of usage.
My line of thinking is:
Claude Code orchestrator -> Local LLM -> Claude Code reviewing the code in the PR.
Anyone tries this? Am I missing something or am I a monday morning genious?
[link] [comments]