LangWatch
langwatch.ai
- Free plan
- Open source
- macOS
- Linux
- Web
- API
- Self-hosted
What LangWatch is
LangWatch is an AI agent engineering platform for teams working to ship reliable agents to production. Its simulations cover multi-turn conversations, tool use, recovery, and business outcomes before launch. Evaluations can run offline against datasets, continuously on production traffic, or in CI/CD to gate merges. Traces show agent steps alongside cost, token use, duration, errors, evaluations, and annotations. Integrations include Python and TypeScript SDKs, OpenTelemetry, REST ingestion, model providers, and frameworks such as LangGraph, LangChain, CrewAI, and DSPy. Voice-agent testing supports OpenAI Realtime, Gemini, ElevenLabs, Google Live API, Pipecat, WebSocket bots, and Twilio. Teams can choose managed cloud, self-hosted, or hybrid deployment; self-hosting uses Docker Compose and the customer's ClickHouse. LangWatch is open source under Apache 2.0. Its trust center lists encryption, access controls, automatic PII detection and removal, and GDPR right-to-be-forgotten support. The free Developer plan includes 50k events per month and two users. Growth is €29/user/mo, billed per core seat.
Who it suits
LangWatch suits AI teams evaluating and monitoring agents, with workflows for developers, product owners, QA, and domain experts. Its deployment choices include managed cloud, self-hosted, and hybrid setups.
Good to know, and what to watch
- Simulations cover tool use and recovery
- Evaluations run on datasets, production, or CI/CD
- Traces report cost, tokens, duration, and errors
- Supports several agent frameworks and voice systems
- Offers managed, self-hosted, and hybrid deployment
- Developer plan includes 14-day data access
- Growth includes 200k events before overage charges
- Self-hosting requires Docker Compose and customer ClickHouse
Plans
Every plan and priceEnterprise
Developer
Growth
Free plan
The verdict
LangWatch brings agent simulations, evaluations, and tracing into one engineering workflow. The free tier is limited in data access and user count, while Growth is priced per core seat.
What the maker publishes
- Agent simulations
- Its simulations test multi-turn agent conversations, tool use, recovery, and business outcomes before launch. source
- Agent simulations
- It runs realistic multi-turn text and voice simulations with simulated users, judges, tool-call verification and visual run debugging. source
- AI gateway
- The AI Gateway provides an OpenAI- and Anthropic-compatible endpoint with virtual keys, budgets, routing and provider fallbacks. source
- AI governance
- Governance features include organization-wide policy controls, audit exports, spend dashboards and inventory of AI tools. source
- CLI and MCP
- The project provides a CLI and an MCP server for use with coding assistants and other MCP clients. source
- Company
- The site's footer identifies the maker as LangWatch B.V. and says it is built in Amsterdam. source
- Company founded
- LinkedIn lists LangWatch as founded in 2023. source
- Company headquarters
- LinkedIn lists headquarters at Herengracht 551, Amsterdam, Netherlands. source
- Company legal name
- Crunchbase identifies the operating legal entity as Reasoning Engine B.V. source
- Company size
- LinkedIn lists the company size as 11–50 employees. source
- Compliance
- LangWatch states ISO 27001 certification, GDPR compliance, EU data residency and SOC 2 coverage through AWS. source
- Deployment
- LangWatch offers managed cloud, self-hosted, and hybrid deployment; its pricing page says self-hosting uses Docker Compose and the customer's ClickHouse. source
- Documentation
- Documentation covers introduction, integrations, CLI, MCP, agent testing, observability, evaluations, datasets, prompt management, self-hosting and API reference. source
- Evaluations
- Evaluations can run offline on datasets, continuously on production traffic, or in CI/CD to gate merges. source
- Evaluations
- The platform supports offline and online evaluations, CI/CD gates, multimodal scoring, built-in evaluators and custom workflows. source
- Framework integrations
- Framework support includes OpenAI Agents, LangGraph, LangChain, CrewAI, Pydantic AI, Agno, Mastra, Vercel AI SDK, Google ADK, LangFlow and Flowise. source
- Funding
- Crunchbase lists LangWatch as a private pre-seed company with two funding rounds and investors including Volta Ventures and Antler. source
- Headquarters
- Amsterdam source
- Integrations
- The integrations page lists Python and TypeScript SDKs, OpenTelemetry and REST ingestion, model providers, and agent frameworks including LangGraph, LangChain, CrewAI, and DSPy. source
- Integrations
- Official integrations cover OpenAI, Anthropic, Google Gemini, Bedrock, Azure OpenAI, Vertex AI, Groq, Ollama and LiteLLM. source
- Intended users
- The company says the platform is for AI teams shipping agents and describes evaluation workflows for developers, product owners, QA, and domain experts. source
- Licence
- The core platform is Apache 2.0; enterprise modules require a commercial production licence and the SDKs are MIT licensed. source
- Limit: review evidence
- G2 says there are insufficient LangWatch reviews to provide buying insight, and Software Advice lists no user reviews. source
- Limit: service reliability
- The public status page shows Scenarios at 96.174% uptime and Langy at 96.823% uptime for the displayed 30-day period. source
Checked against the maker's pages on 21 Oct 2026.