W&B Weave

What it does

W&B Weave takes in traces from agent and AI application runs, including sessions, turns, steps, tools, and sub-agents, along with the inputs, outputs, and metadata for each inference. It produces monitoring views, built-in and custom signals that classify agent interactions, alerts routed to Slack or webhooks, evaluation comparisons, leaderboards, and safety and quality scores for toxicity, bias, PII, hallucinations, coherence, and context relevance.

People use Weave to trace multi-agent behavior and find root causes of failures, then run evaluations to check that changes to prompts, harnesses, or models do not cause regressions. A Playground lets teams test new or custom models against production traces. Coding agents such as Claude Code can connect through the MCP server to read production data, run evaluations, and iterate automatically.

Weave suits developers and teams building agentic or GenAI applications. Access is through the CoreWeave sign-up and Weave docs, with cloud pricing under CoreWeave Forge; the Pro plan starts at $60 per month billed monthly, with a 30-day free trial. Limits include a 1 GB or 1.5 GB monthly data ingestion allowance on lower tiers and pay-per-use overage charges, and Enterprise options are custom-priced.

What it can do

• End-to-end observability of agent sessions, turns, steps, tools, and sub-agents
• Built-in and custom behavior signals that automatically classify agent interactions
• Alerts routed to Slack notifications and webhook automations
• Flexible imperative evaluation API with comparisons and visualizations for regression checks
• Playground for testing prompts and models against production traces
• Guardrail scorers for toxicity, bias, PII detection, hallucinations, coherence, fluency, and context relevance

Works with

• Slack
• Webhooks
• MCP server (connects coding agents such as Claude Code)

Example workflows

Diagnose a production failure

Illustrative example. Input: production agent traces. Steps: open a session in Weave Traces, review turns and tool calls, apply out-of-box signals to flag failing interactions, route an alert to Slack. Output: a pinpointed failure mode and its supporting trace view.

Check a prompt or model change for regressions

Illustrative example. Input: a candidate prompt or model and an evaluation dataset. Steps: run the imperative evaluation API against both versions, compare results in Weave's evaluation visualizations, and review guardrail scores. Output: a comparison showing whether the change improved or regressed reliability before release.

Before you choose

• Cloud pricing is set by CoreWeave Forge and is distinct from W&B privately hosted license pricing; the two should not be compared directly.
• Lower tiers include a monthly data ingestion allowance; overage is charged per MB, so trace volume affects cost.
• Confirm private-deployment support and licensing specifically for Weave; general W&B personal-server terms are not assumed to apply.

Listed under

• Coding
• DevOps and observability
• Agent evaluation
• Agent observability
• AI evaluation
• Developers
• Guardrails
• LLM tracing

Get started

Quickstart and free trial are linked from the product page.

1. Sign up for a CoreWeave account at the sign-up link on the Weave page.
2. Follow the Weave quickstart in the docs to instrument your agent or application with tracing.
3. Optionally connect a coding agent via the MCP server and W&B skills to let it read production data and run evaluations.

Forge Free provides personal-development access. Pro starts at $60/month billed monthly with a 30-day trial and is limited to eligible early-stage teams under 50 employees. Trace ingestion and storage have allowances and overage; consult the Weave rows of the current rate table. Private deployments require separate confirmation.

30-day free trial.

Source: coreweave.com/forge-pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• W&B Weave: Observability for production agents​ ↗wandb.ai
• CoreWeave Forge Pricing ↗www.coreweave.com
• Explore Weights & Biases pricing plans ↗wandb.ai

Company website

wandb.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.