LangWatch Releases Entire AI Agent Observability Platform as Open Source

Amsterdam based LangWatch.ai has released its full AI agent observability and testing platform as open source under the Apache 2.0 license. The release includes all major components of its commercial system such as tracing, evaluations, prompt management, datasets, experiments, the AI gateway, and Scenario, its agent simulation and testing framework.

The entire platform can now be self-hosted using a single command to bring up a local instance. This allows teams to trace agent behavior, inspect unexpected outcomes, manage costs through the gateway, and run simulations. Commercial cloud and enterprise features, including single sign-on, audit logs, and advanced governance, remain part of LangWatch’s paid offerings.

The company also introduced three updates alongside the open source release. Scenario now includes native voice testing that supports multiple real voice systems and can run headlessly in continuous integration environments. The Traces UI has been redesigned to display multi-turn and multi-agent flows, and includes an assistant for investigating errors. A new AI Governance beta provides developers and administrators with visibility into cost attribution, anomaly detection, and audit logs, with all functions available through API, CLI, and MCP server interfaces.

LangWatch stated that the release aims to make AI agent monitoring and testing tools accessible for teams operating in regulated and sensitive environments.