Run agent
Execute prompts on Ollama from the dashboard. Every run is traced automatically with zero API cost.
intelligence + performance
Everything you need to train local agents
AITracer is built around a simple loop: run locally, trace everything, coach when you need quality, export when you are ready to train.
Coach mode
Local agent drafts first; optionally send to a paid model for critique. You pick which version becomes training data.
Trace library
Search prompt/response history, filter by action name, and inspect tokens, latency, and cost per run.
Training export
Download OpenAI or Alpaca JSONL from traces, or build a custom Ollama model with examples baked in.