Helicone
Open-source LLM observability and AI gateway in one line of code.

Open-source LLM observability and evaluation
Langfuse is our top recommendation for teams that want LLM observability without vendor lock-in. Moving essentially the entire product to MIT was a rare, genuinely developer-friendly move: you can self-host the full feature set for free, with no seat or usage caps. The tracing, prompt management, and evaluation tooling are solid and framework-agnostic, so it fits whether or not you use LangChain. The ClickHouse acquisition looks like a stable, aligned home. If open source and self-hosting matter, start here.
Langfuse is an open-source LLM observability platform providing tracing, prompt management, evaluations, datasets, and analytics. It is framework-agnostic and can be fully self-hosted for free (nearly all features are MIT-licensed) or used as a managed cloud service. It was acquired by ClickHouse in January 2026.
Langfuse is an open-source engineering platform for observing, debugging, and improving LLM applications. It captures detailed traces of LLM calls, chains, and agent steps, so teams can see exactly what happened in a request, including inputs, outputs, latency, token usage, and cost. On top of tracing it provides prompt management with versioning, evaluations (including LLM-as-judge and human annotation), datasets, a playground, and analytics dashboards. A key part of Langfuse's appeal is that it is framework-agnostic and unusually generous with open source. In mid-2025 the company moved effectively all product features to the MIT license, leaving only thin enterprise compliance features (such as SCIM, audit logs, and fine-grained RBAC) as commercial. This makes self-hosting the full product genuinely viable, with no seat caps, retention limits, or usage caps, which distinguishes it from many observability tools whose open-source editions are heavily gated. Langfuse offers a managed cloud with a free tier and paid Core and Pro plans, plus self-hosting for free and an enterprise self-host license. It was acquired by ClickHouse in January 2026, a natural fit given Langfuse already used ClickHouse for its trace storage. It competes with LangSmith and other LLM observability tools, differentiating on open-source depth and self-hosting freedom.
Langfuse is an open-source LLM observability platform with tracing, prompt management, evaluations, datasets, and analytics. Nearly all features are MIT-licensed, so the full product can be self-hosted for free with no seat or usage caps, and a managed cloud with free and paid tiers is available. It is framework-agnostic, working with or without LangChain. It was acquired by ClickHouse in January 2026.
Langfuse was founded by Clemens Rawert, Marc Klingen, and Max Deichmann and went through Y Combinator (W23). It built an open-source LLM engineering platform with a strong community following.
Langfuse raised a $4 million seed round and, in January 2026, was acquired by ClickHouse, the analytics database company, a strategic fit because Langfuse's trace storage was already built on ClickHouse. It continues to operate as an open-source project.
Langfuse captures detailed traces of LLM calls, chains, and agents (inputs, outputs, latency, tokens, cost) and layers on prompt management with versioning, evaluations (LLM-as-judge and human annotation), datasets, a playground, and analytics dashboards.
It is framework-agnostic via SDKs (Python/JS) and OpenTelemetry, integrating with LangChain, LlamaIndex, the Vercel AI SDK, and more. Deployment options are managed cloud or free self-hosting of the MIT-licensed full product, with enterprise compliance features available commercially.
Engineering and AI teams that need observability, evaluation, and prompt management for LLM applications, especially those that value open source and self-hosting.
AI and backend engineers instrumenting and debugging LLM apps.
Engineering leaders adopting an observability and evaluation stack, often with self-hosting needs.
Open-source advocates, MLOps engineers, and the AI developer community.
Technical teams that want deep, open-source LLM observability with the freedom to self-host the full product.
Raised a $4 million seed round; acquired by ClickHouse in January 2026. Verify with the vendor.
Yes, and unusually so. In mid-2025 nearly all product features moved to the MIT license, so you can self-host the full platform for free, with only thin enterprise compliance features remaining commercial.
Yes. Self-hosting the MIT version gives the full feature set with no seat caps, retention limits, or usage caps, which is unusually generous among observability tools.
Yes. Langfuse is framework-agnostic and integrates via its SDKs and OpenTelemetry, working with LangChain, LlamaIndex, or plain API-based apps.
Langfuse was acquired by ClickHouse in January 2026, a natural fit since Langfuse already used ClickHouse for trace storage. It continues as an open-source project.
Both provide LLM tracing and evaluation. Langfuse is open source with a strong self-hosting story and no usage caps on the MIT version, while LangSmith is tightly integrated with the LangChain ecosystem. Choose based on hosting and framework preferences.
Side-by-side pages for pricing, features, and best-fit use cases.
Open-source LLM observability and AI gateway in one line of code.
The AI developer platform for experiment tracking and LLMOps.
The open hub for machine learning models, datasets, and demos.
Very fast LLM inference on custom LPU hardware.