Both trace agents, run evaluations on datasets, score production traffic, queue traces for review and version prompts. LangSmith fits teams building on LangChain, LangGraph or Deep Agents; W&B Weave, now part of CoreWeave Forge, fits teams that already train or track models on Weights & Biases.[1][6]
LangSmith is LangChain's platform for tracing, evaluating and versioning prompts for agents, priced per seat plus traces; W&B Weave is Weights & Biases' tracing and evaluation toolkit, sold inside CoreWeave Forge since 30 September 2026 after CoreWeave acquired W&B in May 2025. Feature for feature they are close. The deciding question is usually which ecosystem you already work in: LangChain or Weights & Biases.
| Decision factor | LangSmith | W&B Weave |
|---|---|---|
| Ownership | LangChain, independent; $125M Series B in October 2025 | CoreWeave, which completed its acquisition of W&B in May 2025; sold in CoreWeave Forge since 30 September 2026 |
| Home ecosystem | LangChain, LangGraph and Deep Agents | W&B Models and CoreWeave Forge |
| SDKs | Python, TypeScript, Go, Java | Python; TypeScript with fewer features |
| Prompt tooling | Prompt Hub with commit-based versioning and a playground | Versioned prompt objects and a Playground |
| Run-time checks | Partial: LLM Gateway (beta) for spend caps, model access and redaction on model calls | Partial: Python guardrail scorers that block or modify inputs and outputs |
| Self-hosting | Enterprise add-on, or a BYOC data plane | Kubernetes, with a W&B Platform install and a Weave-enabled licence |
| Free tier | Developer: 1 seat, 5k base traces a month | Forge Free: up to 5 model seats, 1 GB Weave ingestion a month |
| First paid tier | Plus: $39 per seat a month, 10k base traces | Forge Pro: from $60 a month, 1.5 GB ingestion, teams under 50 people |
| Capability | LangSmith | 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 | Partial | ✓ |
| Gateway or proxy in the request path | — | Not stated | — |
| Licence and deployment | |||
| Open source | Partial | Partial | Partial |
| Self-hosting | ✓ | ✓ | Partial |
CoreWeave completed its acquisition of Weights & Biases on 5 May 2025, and on 30 September 2026 launched CoreWeave Forge, which bundles W&B Models, Weave and newer CoreWeave products. Hosted W&B accounts now sign in through Forge; self-hosted and dedicated-instance accounts are unchanged. Weave's docs, pricing and login now live on CoreWeave domains.
Inside Weave, the agent story is shifting. CoreWeave recommends Agent Lens, a separate product in public preview that shares Weave's trace backend, for agent tracing and analysis, and Weave's docs call its built-in monitors a previous approach. LangSmith's packaging changed this month too: self-serve customers moved to new plans, with several products metered in LangChain Standard Units.
LangSmith has the deepest integration with LangChain, LangGraph and Deep Agents and also lists the OpenAI Agents SDK, Claude Agent SDK, CrewAI, Google ADK, Mastra, PydanticAI and others. Beyond tracing, it covers agent deployment (formerly LangGraph Platform) and Engine, in public beta, which reads production traces and drafts fixes. SDKs cover Python, TypeScript, Go and Java at similar depth.
Weave is strongest for Python teams already in Weights & Biases. You trace with the @weave.op decorator or provider auto-patching, run the Evaluation API over datasets with LLM-judge, function or class-based scorers, and publish leaderboards. Prompts can be linked to the CoreWeave Registry, and the Playground can route to CoreWeave's inference. The TypeScript SDK lacks class-based scorers, guardrails and cost tracking, so mixed-language teams should check that gap.
Evaluation features line up closely. LangSmith runs LLM-as-judge, pairwise and code evaluators offline and online, with online evaluators triggered by automation rules that filter and sample live traffic, including multi-turn threads. Its annotation queues support single-run and pairwise review with rubrics and assigned reviewers. Weave's monitors score incoming production traces with LLM-judge signals and send alerts through Slack or webhooks, and its annotation queues take boolean, numeric, categorical or free-text feedback that can be exported to datasets. Both track cost: LangSmith covers listed providers automatically and accepts manual costs for others, while Weave's cost tracking works from Python only.
Both accept OpenTelemetry and read the OpenInference and OpenLLMetry conventions, so a team that instruments with OpenTelemetry can try both without rewriting its tracing.
The two take different routes to acting during a run. LangSmith's LLM Gateway, in public beta since 30 July 2026, sits in the model-call path when you change the provider base URL. It enforces spend caps, model access policies and redaction of personal data and secrets, with custom regex added in October 2026. It does not cover tool calls; tool and MCP gateways are described as planned.
Weave's guardrails are scorers you call from Python application code. They can block or modify inputs and outputs before they reach users, which means code changes and no TypeScript support. Neither documents approvals or policy enforcement on an agent's tool actions.
As of 20 October 2026, LangSmith's Developer tier is free for one seat and 5k base traces a month with 14-day retention, and Plus is $39 per seat a month with 10k base traces. The per-trace overage rate is not published. Extended retention on LangSmith's hosted service is capped at 180 days.
Weave is billed through CoreWeave Forge plans that combine model seats, storage and Weave ingestion. Forge Free includes up to five model seats and 1 GB of Weave ingestion a month. Forge Pro starts at $60 a month with ten seats and 1.5 GB of ingestion, is limited to teams under 50 people, and lists overage at $0.10 per MB; Enterprise is quoted, with overage listed at $0.20 per MB. LangSmith's bill grows with headcount and trace count; Weave's with data volume. Check LangSmith's pricing and CoreWeave Forge pricing.
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 and a kill switch for a run, an agent or a team. LangSmith's gateway covers model calls and Weave's guardrails check content from Python code; neither holds a tool action for approval. Prefactor has no prompt playground or prompt registry and its platform is hosted, so if prompt iteration is central, LangSmith or Weave 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.