Agenta is an open-source LLMOps platform that provides a unified workspace for prompt engineering, versioning, evaluation, and production observability of LLM applications. It supports 50+ LLM providers and frameworks including LangChain and LlamaIndex without vendor lock-in. Teams can run automated evaluations (LLM-as-judge, custom evaluators) alongside human annotation workflows, and track all prompt changes with branching and environment-based rollouts. It is available as a managed cloud service or self-hosted. Key features: - Interactive LLM playground with side-by-side prompt comparison - Prompt version control with branching, environments, and rollback - Automated and human-in-the-loop evaluation workflows - Production observability: tracing, latency, cost, and error tracking - Model-agnostic: 50+ providers, works with LangChain and LlamaIndex - Self-hosted or managed cloud; open source on GitHub
Hobby: free (2 users, 5k traces/month, 20 evals/month). Pro: $49/month (3 users, 10k traces, unlimited evals). Business: $399/month (unlimited seats, 1M traces). Enterprise: custom pricing.