LangWatch is an Amsterdam-based LLMOps platform founded in 2023 by Manouk Draisma and Rogério Chaves (alumni of Booking.com and Lightspeed). It provides a developer-first but cross-functional environment for defining evaluations, running experiments, simulating multi-step agent behavior, and monitoring LLM applications in production. The platform is open-source with over 3,000 GitHub stars and raised a €1 million pre-Seed round in February 2025 led by Passion Capital, with participation from Volta Ventures and Antler. The core LLM observability layer is built on OpenTelemetry, enabling trace-level inspection of prompts, tool calls, and agent decisions across environments without vendor lock-in. It integrates with all major LLM frameworks and providers and supports cost monitoring across 800+ models, providing customizable dashboards for usage trends and spend optimization. LangWatch includes an end-to-end agent simulation capability that runs realistic scenarios against the full stack (tools, state, user simulator, and an LLM judge), pinpointing exactly where and why agents fail. Its evaluation system lets teams create and tune custom evals that measure product-specific quality dimensions, and it supports DSPy-based prompt optimization. An AI Gateway component acts as an OpenAI/Anthropic-compatible proxy with virtual keys, hierarchical budgets, inline guardrails, automatic provider fallback, and Anthropic cache_control passthrough for governance and cost control. The platform is ISO 27001 and SOC 2 certified, GDPR compliant, and supports self-hosted deployment via Docker Compose, Kubernetes (Helm), or on-prem setups on AWS, GCP, and Azure. Key features: - LLM observability via OpenTelemetry with full trace inspection of prompts, tool calls, and agent steps - End-to-end AI agent simulation with user simulator and LLM judge for pre-production testing - Custom evaluation framework with real-time eval execution and DSPy prompt optimization - AI Gateway: OpenAI/Anthropic-compatible proxy with virtual keys, budgets, guardrails, and provider fallback - Cost monitoring and analytics across 800+ models and providers - Prompt and model version management with feature-flag-style rollout and audit trails - ISO 27001, SOC 2 certified and GDPR compliant; supports on-prem and self-hosted deployment - Open-source core with unlimited lite seats for stakeholder visibility
Free Developer plan (no credit card required). Paid plans from €59/month (~$65) including unlimited evaluations and DSPy optimization. Seats priced at €29/seat/month with unlimited read-only lite seats. Enterprise plans with on-prem deployment available on request.