Galileo
Evaluation and observability platform that turns tuned evals into low-cost runtime guardrails for agents
Galileo is an AI observability and evaluation platform, now part of Cisco, built to turn offline evaluations into production guardrails. Teams build datasets from synthetic, development and live data, tune evals to their environment, then run those evals on live traffic to stop agent failures instead of only reporting them.
Key features:
- Start with more than 20 built-in evals for RAG, agents, safety and security.
- Write custom evaluators that encode domain expertise.
- Auto-tune metrics from live feedback and expert annotations.
- Distill LLM-as-judge evaluators into compact Luna models that monitor all traffic at low cost.
- Find failure modes and get suggested fixes from an insights engine.
- Bring unit testing and CI/CD practices to AI development.
Galileo can run as SaaS, in a VPC or on premises. Pricing is available on request, and access starts with a demo; the pricing page now redirects to Splunk.