LangWatch is an open-source agent observability and evaluation platform for AI platform teams, used to trace, test, route, and govern every LLM call in production, covering both the agents teams build and the coding assistants engineers use. The project is licensed under Apache 2.0, with enterprise modules under the enterprise/ directory requiring a commercial license, and the SDK under the MIT license. LangWatch Cloud and self-hosted options are officially offered.
The main uses and capabilities fall into four categories. LLM Ops: observability, agent testing, evaluation, and prompt management. Coding agents: providing session, per-pull-request, and team cost tracking, as well as privacy controls. AI Gateway: providing unified OpenAI- and Anthropic-compatible endpoints, virtual keys, budgets, and routing. AI governance: allowing visibility into all AI tools and users across the company, with support for anomaly rules.
Examples of use cases: you can create simulated tests for evaluating or benchmarking agents and voice agents; measure the cost of each PR and optimize Claude Code usage to extend the weekly limit; create virtual keys with budgets to distribute to customers or employees; and discover that 40% of Copilot Studio subscriptions in the company are not actually used and cancel the subscriptions.
Supported coding assistants include Claude Code, Codex, GitHub Copilot, opencode, Cursor, MCP server, and others; frameworks include LangChain, LangGraph, Vercel AI SDK, Mastra, CrewAI, Google ADK, DSPy, OpenTelemetry, and others; model providers include OpenAI, Anthropic, Azure OpenAI, Vertex AI, Bedrock, and others; governance sources include Copilot Studio, Claude Cowork, the Anthropic and OpenAI compliance APIs, Workato, Databricks Genie, OpenTelemetry, and S3 audit sources, and others; no-code platforms include LangFlow, Flowise, n8n, and others.
For installation, the official instructions state that Node.js is required. For self-hosting LangWatch, see the documentation. To contribute, you can read CONTRIBUTING.md and run quickstart; bugs and feature requests are submitted to GitHub issues, and questions can be discussed on Discord. LangWatch is GDPR compliant and ISO 27001 certified, and security vulnerabilities can be reported to security@langwatch.ai. More detailed installation commands, configuration steps, and pricing information are not currently provided.
