I feel like I'm always behind the LLM crazy-curve. I'm also getting sick of context switching between LLM apps. Me and Claude Code are basically down to pair-writing test cases, documentation, and commit messages. Now that Perplexity is moving into the desktop space I thought I'd show what I've got to HN and see if I can get some more hands on King Louie.
It has 13 providers, rules to pick the best LLM for each task, agents with a couple dozen built-in tools, semantic memory with embeddings, P2P mesh networking, and a skill system. The LLM-written README.md gives more in-depth details.
Happy to answer questions about the architecture, routing engine, and how I deal with context reduction. Also actively looking for contributors. Cross platform, JavaScript/Electron, MIT licensed, and plenty of unsolved problems left if you're into agent tooling, LLM routing, Context Reduction, or P2P networking.