Both trace agents, run experiments, score production traffic, queue traces for human review and version prompts. Langfuse is the open-source option you can self-host for free; Weave is the choice when your model training already lives in Weights & Biases or on CoreWeave. Both changed owners, and both say their products continue.[3][6]
The difference between Langfuse and W&B Weave is mostly how you can run them: Langfuse is MIT-licensed and free to self-host, while self-hosting Weave needs a W&B licence and an existing W&B Platform install. Both changed owners: ClickHouse acquired Langfuse in January 2026, and CoreWeave, which bought W&B in May 2025, has sold Weave inside CoreWeave Forge since September 2026. Weave can block content during a run through scorers in Python code; Langfuse evaluates after the fact.
| Decision factor | Langfuse | W&B Weave |
|---|---|---|
| Ownership | Acquired by ClickHouse in January 2026; licence and self-hosting unchanged, per Langfuse | CoreWeave since May 2025; sold inside CoreWeave Forge since 30 September 2026 |
| Licence | MIT, except add-ons in the ee folders | Repository Apache-2.0; self-hosting needs a Weave-enabled licence |
| Self-hosting | Free, via Docker or Kubernetes | Kubernetes, on top of a licensed W&B Platform install |
| TypeScript | Python and TypeScript SDKs | TypeScript SDK lacks guardrails, class-based scorers and cost tracking |
| Runtime guardrails | None; evaluation happens after the fact | Partial: scorers called from Python code can block or modify inputs and outputs |
| Online evals | Evaluator rules with sampling, plus backfills over past data | Monitors on production traces; docs now recommend Signals |
| Free hosted tier | Hobby: 50k units a month, 30 days of data | Forge Free: 1 GB of Weave ingestion a month |
| First paid tier | Core: $29 a month, 100k units | Forge Pro: from $60 a month, 1.5 GB of Weave ingestion |
| Capability | Langfuse | W&B Weave | 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 | — | Partial | ✓ |
| Gateway or proxy in the request path | — | Not stated | — |
| Licence and deployment | |||
| Open source | ✓ | Partial | Partial |
| Self-hosting | ✓ | ✓ | Partial |
Langfuse joined ClickHouse in January 2026 and said its roadmap and licence would not change and that self-hosting would stay a first-class option. It has kept shipping as its own product: experiments rebuilt in April, code evaluators in May, a 1.0 command-line tool in August and evaluator backfills in September. Langfuse Cloud becomes v4-only on 16 November 2026, which may mean migration work for existing cloud users.
Weave's owner change came earlier and goes further. CoreWeave completed its acquisition of Weights & Biases in May 2025, and on 30 September 2026 launched CoreWeave Forge, which bundles W&B Models, Weave, Agent Lens and CoreWeave inference. W&B hosted accounts now sign in through Forge, and Weave's pricing, docs and login moved to CoreWeave domains. CoreWeave now recommends Agent Lens, a separate product in public preview, for agent tracing. Weave is less a standalone product than one part of a model platform, which is its main strength for W&B users and its main cost for everyone else.
Both self-hosted stacks use ClickHouse and object storage, so the operating work is similar. The difference is the licence. Langfuse publishes its platform under MIT and documents free self-hosting on Docker Compose or Kubernetes, backed by Postgres, ClickHouse, Redis and S3-compatible storage; a few add-on features in its ee folders need a licence key.
Weave's code is in an Apache-2.0 repository, but a self-hosted deployment needs a Weave-enabled licence from W&B and an existing W&B Platform installation, on Kubernetes 1.29 or later with ClickHouse and S3-compatible storage. If you already self-host W&B for training, adding Weave may be straightforward. If you do not, Langfuse is the option that works without a sales conversation.
Langfuse offers Python and TypeScript SDKs, with code evaluators in either language since May 2026, and accepts OTLP over HTTP from any language. Weave's SDK centres on a Python decorator; its TypeScript SDK does not support guardrails, class-based scorers or cost tracking. Both accept OTLP over HTTP only, and both read OpenInference-style attributes. A TypeScript-heavy team will find Langfuse more complete.
Evaluation features are close. Langfuse's evaluator rules score live production observations with sampling, and since September 2026 can backfill scores over up to six months of past data; a GitHub Action fails a CI run below a score threshold. Weave scores production traces with monitors and alerts through Slack or webhooks, though its docs now call built-in monitors a previous approach and recommend Signals. Both have annotation queues, versioned prompts and a playground.
The run-time picture reverses for Python teams. Weave's guardrails are scorers called from Python application code that can block or modify inputs and outputs before they reach users. Langfuse does not block anything at run time; its documentation says protection comes from separate libraries and Langfuse evaluates those afterwards, with threshold alerts that notify rather than stop. Neither documents approvals on an agent's tool calls.
As of 20 October 2026, Langfuse prices per unit ingested, counting traces, observations and scores: Hobby is free for 50k units a month with 30 days of data, Core is $29 a month for 100k units, then $8 per 100k, and Pro is $199 a month. CoreWeave prices Weave through Forge by seats, storage and ingestion: Forge Free includes 1 GB of Weave ingestion a month and up to 5 model seats, and Forge Pro starts at $60 a month with 1.5 GB for teams under 50 employees, with extra ingestion listed at $0.10 per MB.
Langfuse counts events and Weave counts bytes, so trace size decides which is cheaper. Self-hosted Langfuse has no subscription at all. At the top end, Langfuse Enterprise is $2,499 a month or custom, and Weave Enterprise is a custom quote. Check Langfuse's pricing and CoreWeave Forge pricing against your own volumes.
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 on tool calls and data access, approval holds for risky actions and a kill switch for a run, an agent or a team. Langfuse evaluates after the fact, and Weave's guardrails check content from Python code without approvals on agent actions. 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 or Weave is the better pick. 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.