Both trace agents, run experiments, score production traffic, queue traces for human review and version prompts. Langfuse is open source with free self-hosting and a low first paid tier; HoneyHive is a hosted platform that sells everything above its free tier through sales, with integrations aimed at enterprise agent platforms.[1][4]
Langfuse is MIT-licensed (apart from licensed add-ons) and free to self-host, with a $29 a month hosted tier; HoneyHive is a proprietary hosted platform with a free Developer tier and quote-based Enterprise plans, including self-hosting. ClickHouse acquired Langfuse in January 2026 and Langfuse says its licence and self-hosting are unchanged; HoneyHive is independent. Both are built on OpenTelemetry and neither blocks live agent actions.
| Decision factor | HoneyHive | Langfuse |
|---|---|---|
| Ownership | Independent; $7.4M total funding last announced April 2025 | Acquired by ClickHouse in January 2026; licence and self-hosting unchanged, per Langfuse |
| Licence | Proprietary platform; SDKs and tools MIT | MIT, except add-on features in the ee folders |
| Self-hosting | Enterprise only; fully self-hosted or hybrid, documented on AWS | Free, via Docker or Kubernetes |
| OpenTelemetry | Built on it; OTel GenAI, OpenLLMetry and OpenInference conventions | OTLP over HTTP; gRPC not supported |
| SDKs | Python tracing SDK; TypeScript REST client | Python and TypeScript |
| Run-time controls | None; Policies set evaluation standards | None; evaluation happens after the fact |
| Free tier | Developer: 10,000 events a month, 30-day retention, up to 5 users | Hobby: 50k units a month, 30 days of data |
| First paid tier | No self-serve paid plan; Enterprise is quote-based | Core: $29 a month, 100k units |
| Capability | HoneyHive | Langfuse | Prefactor |
|---|---|---|---|
| Observe | |||
| LLM call tracing | ✓ | ✓ | ✓ |
| Multi-step agent tracing (tool calls, nested spans) | ✓ | ✓ | ✓ |
| Cost tracking | ✓ | ✓ | ✓ |
| OpenTelemetry support | ✓ | ✓ | ✓ |
| Evaluate | |||
| Offline evals (datasets, experiments) | ✓ | ✓ | ✓ |
| Online evals on production traffic | ✓ | ✓ | ✓ |
| LLM-as-a-judge scoring | ✓ | ✓ | ✓ |
| Code-based evaluators | ✓ | ✓ | ✓ |
| Human review and annotation | ✓ | ✓ | ✓ |
| Prompt versioning and playground | ✓ | ✓ | — |
| Act on findings | |||
| Runtime controls on live agent actions | — | — | ✓ |
| Gateway or proxy in the request path | — | — | — |
| Licence and deployment | |||
| Open source | Partial | ✓ | Partial |
| Self-hosting | ✓ | ✓ | Partial |
Langfuse publishes its platform under the MIT licence, with a small set of add-on features in licence-keyed ee folders. Self-hosting is free on Docker Compose for testing or Kubernetes for production, backed by Postgres, ClickHouse, Redis and S3-compatible storage. ClickHouse acquired Langfuse in January 2026; Langfuse said its roadmap would not change, there were no planned licensing changes and self-hosting would stay a first-class path.
HoneyHive's platform is proprietary; only its SDKs, CLI and related tooling are MIT-licensed. Self-hosting comes with an Enterprise contract, either fully self-hosted or as a hybrid where HoneyHive runs the control plane, and is documented on AWS. If you need to run the software yourself without a sales process, Langfuse is the one that allows it.
The size of the open project differs too. The Langfuse repository had about 36,000 GitHub stars when checked on 20 October 2026 and published release v4.64.0 the day before. HoneyHive's public repositories hold SDKs and tooling, so there is no server code to inspect or fork. For teams that want to audit or patch the platform themselves, that matters more than the feature lists.
As of 20 October 2026, Langfuse's hosted Hobby tier is free for 50k units a month (traces, observations and scores) with 30 days of data. Core is $29 a month for 100k units, then $8 per 100k, with 90 days of data. Pro is $199 a month with three years of data access, and Enterprise is $2,499 a month or custom.
HoneyHive's Developer tier is free with 10,000 events a month, 30-day retention, up to five users and one workspace. There is no published paid tier; anything beyond that is an Enterprise quote, which adds unlimited users and workspaces, SSO, data residency and dedicated, hybrid or self-hosted deployment. A team expecting to grow past the free limits in a few weeks should factor in a procurement step with HoneyHive. See HoneyHive's pricing and Langfuse's pricing.
Both list SOC 2 Type II. Langfuse also lists ISO 27001 and a HIPAA-ready cloud region, with a BAA on Pro and Enterprise; HoneyHive offers a BAA and EU data residency on Enterprise.
Both accept OpenTelemetry, so instrumentation can move between them. HoneyHive reads the OTel GenAI, OpenLLMetry and OpenInference conventions and supports bring-your-own instrumentors and the OpenTelemetry Collector. Langfuse accepts OTLP over HTTP as JSON or protobuf, not gRPC, names OpenLLMetry and OpenLIT as compatible and maps OpenInference-style attributes.
Langfuse has Python and TypeScript SDKs and integrations for LangChain, LangGraph, LlamaIndex, the OpenAI Agents SDK, the Vercel AI SDK and others. HoneyHive's full tracing SDK is Python, and its TypeScript package is a REST client, so TypeScript teams lean on OTLP. HoneyHive goes further into enterprise agent platforms, tracing Salesforce Agentforce, Databricks Agent Bricks and Microsoft Copilot Studio, and into coding agents, with session traces and spend analysis for Claude Code and Devin.
Langfuse rebuilt experiments as a top-level feature in April 2026, runnable on datasets, production traces or local data, then added code evaluators and a GitHub Action for CI in May and evaluator backfills over up to six months of past data in September. Langfuse Cloud becomes v4-only on 16 November 2026, which may mean migration work for existing users.
HoneyHive runs experiments with evaluate() and gates pull requests through GitHub Actions. Its online evaluations sample production traces, at 10% by default since May 2026, and annotation queues gained automations in April. Its Policies feature sets evaluation and monitoring standards across an organisation, which is aimed at larger teams. Neither product blocks or holds live agent actions: Langfuse's docs say run-time protection comes from separate libraries, and HoneyHive describes itself as the observability layer in Microsoft's Agent Control Specification partnership.
We sell the layer this section describes. Read it with that in mind.
Stated plainly, since Prefactor publishes this page: Prefactor is an agent evaluation platform that scores each step of a run and acts on what it finds through runtime policies, approval holds and a kill switch scoped to a run, an agent or a team. Neither HoneyHive nor Langfuse documents run-time controls. Prefactor has no prompt playground or prompt registry and its platform is hosted, so if free self-hosting or prompt iteration is what you need, Langfuse is the better fit. See the full comparison of 15 platforms.
Reviewed against public product and documentation pages on October 20, 2026. If a vendor has changed a feature, product name, or positioning since then, send a correction and we will update it. Numbered source links in the page body point to the ordered sources below.
Prefactor scores each step of every agent run and can hold or stop an action when a score says it should not happen. Start on the free tier: 5,000 spans a month, no credit card required.
Prefactor helps teams observe, evaluate, and improve their AI agents in production, across every framework and provider.