The first local-first AI assistant that treats strategy evolution as a first-class concern. Build trust in your agent, one decision at a time.
Lydia solves the "AI Drift" problem through controlled, verifiable growth.
Candidate strategies must pass through 4 gates: Syntax, Risk, Replay Performance, and Human Review.
Explore new behaviors in isolated branches. Merge only what works and what you trust.
Your data, your models, your safety. Run locally with support for Anthropic, OpenAI, and Ollama.
Constant thinking, iterative actions, and seamless observation.
Analyze intent & skills
LLM reasoning loop
Execute MCP tools
Report & Refine
Get up and running in seconds.