LangWatch

LangWatch is a platform for testing, evaluating and observing AI agents: it simulates users against an agent before release, traces every step in production including tool calls and MCP servers, runs evaluations both offline and on live traffic, and red-teams agents with adversarial simulations. Prompt versioning with history and GitHub sync, an AI gateway for routing model calls, and voice-agent testing sit alongside that. It is aimed at engineering and quality teams shipping agents they have to defend, rather than at individual experimentation.

It suits European enterprises for a specific, checkable reason: the site lists EU data residency alongside US, UK and APAC regions, offers self-hosted deployment on Docker, Kubernetes/Helm or in your own VPC plus a hybrid split, and states ISO 27001 certification, GDPR compliance and Vanta monitoring. That is the set of claims a procurement review asks for in one place. It suits less well a small team that simply wants cheap tracing: the deployment options and security controls on offer are enterprise-shaped, and the free entry point is described as a five-minute start rather than a complete free tier.

Not verified: no agent or trace was run here, so the quality of the judge, the simulations and the red-teaming were not observed. The pricing page was not opened, so what the free start includes and where limits fall are unknown. Residency, the certifications and the security controls are the vendor's statements on its own page, not documents read here; the page says the product is open source under Apache 2, but the repository, its activity and what exactly is covered by that licence were not checked.