Both trace agents, run experiments on datasets, score production traces with LLM judges and version prompts. Datadog fits teams that already run Datadog for application and infrastructure monitoring; Weave fits teams whose model work already lives in Weights & Biases or on CoreWeave. The platform you already pay for usually decides this one.[1][6]
The difference between Datadog and W&B Weave is which platform they extend: Datadog Agent Observability, formerly LLM Observability, puts agent traces next to Datadog's APM and infrastructure data, while Weave puts them next to Weights & Biases model training. CoreWeave acquired W&B in May 2025 and has sold Weave inside CoreWeave Forge since September 2026. Datadog is SaaS only; Weave can be self-hosted with a W&B licence.
| Decision factor | Datadog | W&B Weave |
|---|---|---|
| Ownership | Datadog, public company; product renamed Agent Observability in June 2026 | CoreWeave since May 2025; sold inside CoreWeave Forge since 30 September 2026 |
| Sits next to | Datadog APM, logs and infrastructure monitoring | W&B model training, registry and CoreWeave inference |
| Self-hosting | None; Datadog-hosted only, not on US government sites | Kubernetes with a Weave licence and a W&B Platform install |
| Runtime guardrails | Not in Agent Observability; AI Guard is a separate product in limited availability | Partial: scorers called from Python code can block or modify inputs and outputs |
| Code evaluators in production | Run in your app and submitted via SDK or API | Custom monitors and scorers; class-based scorers Python only |
| SDKs | Python, Node.js, Java; Go experimental | Python and TypeScript; TypeScript lacks guardrails and cost tracking |
| Free tier | 40K LLM spans a month, 15-day retention | Forge Free: 1 GB of Weave ingestion a month |
| First paid tier | Pro: $160 a month billed annually for 100K LLM spans | Forge Pro: from $60 a month, 1.5 GB of Weave ingestion |
| Capability | Datadog | 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 |
Datadog Agent Observability, renamed from LLM Observability in June 2026, is one module of the Datadog platform and needs a Datadog account. Its strength is correlation: an agent's trace sits next to the backend services, logs and infrastructure it touched, so an on-call engineer can follow a slow or failing run into the rest of the system.
Weave is the LLM and agent toolkit from Weights & Biases. CoreWeave completed its acquisition of W&B in May 2025, and since 30 September 2026 Weave has been sold inside CoreWeave Forge alongside W&B Models, a registry and CoreWeave inference. W&B hosted accounts now sign in through Forge, and CoreWeave recommends Agent Lens, a separate product in public preview, for agent tracing. Weave's strength is the link to model work: prompts can be linked to the CoreWeave Registry, and its playground can route to CoreWeave's dedicated inference.
Both also track cost. Datadog calculates per-request cost from token counts and public rates for more than 800 text models, with custom values for self-hosted ones. Weave calculates token costs for supported providers, including cache tokens, but only from its Python SDK.
If you run neither platform, both are heavier choices than a standalone evaluation tool, because each makes most sense alongside the rest of its parent.
Datadog documents no self-hosted deployment. Traces go to a Datadog-hosted site, and Agent Observability is not available on Datadog's US government sites. Lapdog, a local trace viewer, runs without an account but is a development tool, not a production option.
Weave offers multi-tenant SaaS, a dedicated single-tenant cloud and a self-hosted deployment on Kubernetes 1.29 or later with ClickHouse and S3-compatible storage. Self-hosted Weave requires a Weave-enabled licence and an existing W&B Platform installation, so it suits teams that already self-host W&B. For an organisation where traces must stay inside its own network, Weave has a documented path and Datadog does not.
The evaluation features are close. Both run experiments against versioned datasets, offer LLM-as-judge evaluators, score sampled production traffic and route traces to human reviewers through annotation queues. One difference sits in code evaluators: Datadog runs them automatically only in experiments, so in production your application runs them and submits the result. Weave runs custom scorers through its monitors.
Language support also differs. Datadog has SDKs for Python, Node.js and Java, with Go experimental, and accepts OTLP directly with a three to five minute ingestion delay. Weave has Python and TypeScript SDKs and accepts OTLP over HTTP; its TypeScript SDK lacks guardrails, class-based scorers and cost tracking.
Neither blocks actions as part of its observability product. Weave's guardrails are scorers you call from Python application code that can block or modify inputs and outputs. Datadog keeps blocking of prompts, responses and tool calls in AI Guard, a separate product in limited availability, billed per token scanned. Neither documents approval workflows for agent actions.
As of 20 October 2026, Datadog bills only LLM spans, the calls to a model provider; tool, agent, workflow and retrieval spans are free. Free covers 40K LLM spans a month with 15-day retention, and Pro is $160 a month billed annually for the first 100K LLM spans, then $3.50 per 10K, with longer retention extra. Weave is priced through CoreWeave Forge by seats, storage and ingestion: Forge Free includes 1 GB of Weave ingestion a month, 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.
An agent that makes many tool calls and few model calls is cheap to trace on Datadog's unit. An agent with large prompts and outputs uses up Weave's byte-based allowance faster. Check Datadog's pricing and CoreWeave Forge pricing with your own traffic.
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. Datadog offers blocking only through AI Guard, a separate product, and Weave's guardrails check content from Python code; neither holds an action for approval. Prefactor is not tied to an APM or training platform, but it has no prompt playground or prompt registry and its platform is hosted, so teams that need those should look at Weave or Datadog. 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.