LangWatch

LangWatch is a platform for testing, evaluating and monitoring AI agents. Its core idea is simulation based testing: simulated users, in text and voice, push an agent through many conversation paths, and a judge agent scores each run and can decide pass or fail at any turn, with reasoning drawn from the whole trace. Scenarios can be written in plain language from an editor such as Claude Code, run locally or in CI, and replayed step by step in a visualizer. Adversarial simulations probe for jailbreaks, policy breaks and unsafe tool calls. It is aimed at engineering teams shipping agents to production. Beyond simulations it offers evaluations that score a single output or a whole conversation, using an LLM judge, custom code or a workflow, from a notebook or the UI, and online evaluation of production traffic. Tool calls, skills and MCP servers are traced, and the platform works with frameworks such as LangGraph, CrewAI and Pydantic AI. The Scenario SDK is open source, and the platform can be self hosted. Pricing is shown in euros, with a dollar toggle. The Developer plan is free forever with 50k events per month, 14 days of data access, a two seat limit and 3 each of scenarios, simulations and custom evals. Growth costs 29 euros per core seat per month, includes 200k events then 5 euros per 100k, 30 days of retention with extra at 3 euros per GB, and unlimited simulations and evals. Enterprise pricing is custom.