Skip to content
SPECIFICTION
Get listed

W&B Weave

site.wandb.ai

From $60/mo Free plan Open source WebWindowsmacOSLinuxiOSAPI
Visit site.wandb.ai
Spec score 8.6 #8 of 31 in AI Agent Observability Tools

What it is

W&B Weave provides observability and continuous-improvement tools for production AI agents. It structures traces as sessions and turns, with steps, tools, and sub-agents represented individually. Built-in and custom signals classify interactions, and alerts can be sent through Slack notifications or webhook automations. Its evaluation tools compare results and visualize regressions before deployment; the Playground lets teams compare models and prompting methods against production traces. Safety scorers cover toxicity, bias, personal information, and hallucinations, while quality scorers assess coherence, fluency, and context relevance. Weave works with LangChain and LlamaIndex and supports models from OpenAI, Meta, Amazon Bedrock, and Anthropic. Python and TypeScript libraries are available, as is OpenTelemetry trace submission through an OTLP endpoint. Weave runs on API, Linux, macOS, self-hosted, web, and Windows platforms. The Free plan costs nothing and includes 1 GB of monthly ingestion and 5 GB of storage. Pro costs $60 a month, with 1.5 GB of monthly ingestion; extra ingestion costs $0.10 per MB. A 30-day trial is listed.

Spec sheet

What it does
W&B Weave provides observability, evaluation and continuous-improvement tooling for production AI agents and LLM applications.Read 20 Oct 2026 · site.wandb.ai
Purpose
W&B Weave provides observability and continuous improvement tools for production AI agents.Read 30 Sep 2026 · site.wandb.ai
Integrations
The deployment guide lists integrations for OpenAI, Anthropic, Cohere, Mistral, LangChain, LlamaIndex, DSPy, Google Gemini, Amazon Bedrock, Together and Groq.Read 20 Oct 2026 · site.wandb.ai
Security
W&B describes AWS, Azure and Google Cloud hosting and says Trust Center certifications and security artifacts are available to approved customers.Read 20 Oct 2026 · wandb.ai
Support
The Pro plan includes priority email and chat support, while Enterprise includes an enterprise support package.Read 30 Sep 2026 · site.wandb.ai
Headquarters
San Francisco, California, United StatesRead 28 Sep 2026 · site.wandb.ai
Founded
2017Read 28 Sep 2026 · site.wandb.ai
Company
Weights & Biases says it was founded in 2017 by Lukas Biewald, Chris Van Pelt, and Shawn Lewis and was acquired by CoreWeave in May 2025.Read 30 Sep 2026 · site.wandb.ai
Acquisition news
CoreWeave announced on 4 March 2025 that it would acquire Weights & Biases and completed the acquisition on 5 May 2025.Read 20 Oct 2026 · investors.coreweave.com
Agent-native tracing
Traces are organized around sessions, turns, steps, tools and sub-agents as first-class objects.Read 20 Oct 2026 · site.wandb.ai
Apple app requirements
The W&B iPhone app requires iOS 18 or later, and its Mac version requires macOS 15 or later on Apple silicon.Read 20 Oct 2026 · apps.apple.com
Behavior monitoring
Built-in and custom signals classify agent interactions, while Slack notifications and webhooks route detected issues.Read 20 Oct 2026 · site.wandb.ai
Company facts
LinkedIn lists Weights & Biases as privately held, founded in 2017, headquartered in San Francisco and employing 201–500 people.Read 20 Oct 2026 · linkedin.com
Deployment options
The deployment guide describes multi-tenant SaaS and single-tenant AWS Dedicated Cloud deployments for Weave.Read 20 Oct 2026 · site.wandb.ai
Documentation depth
The support center lists a dedicated Weave section with 16 articles and eight tags.Read 20 Oct 2026 · docs.wandb.ai
Encryption
W&B states that data is encrypted in transit using TLS 1.2+ and at rest using AES 256.Read 30 Sep 2026 · site.wandb.ai
Enterprise controls
Enterprise pricing lists options including single sign-on, audit logs, secure private connectivity, and customer-managed encryption keys.Read 30 Sep 2026 · site.wandb.ai
Evaluations
An imperative evaluation API compares application or model changes with visualizations intended to expose regressions.Read 20 Oct 2026 · site.wandb.ai
Guardrails
Prebuilt safety scorers cover toxicity, bias, PII and hallucinations, while quality scorers cover coherence, fluency and context relevance.Read 20 Oct 2026 · site.wandb.ai
Leaderboards
Evaluation results can be aggregated into leaderboards and shared across an organization.Read 20 Oct 2026 · site.wandb.ai
Legal name and ownership
Crunchbase identifies the legal entity as Weights & Biases, Inc., founded by Chris Van Pelt, Lukas Biewald and Shawn Lewis, and records seven funding rounds and acquisition by CoreWeave.Read 20 Oct 2026 · crunchbase.com
Licence
The public Weave repository is released under the Apache-2.0 license.Read 20 Oct 2026 · github.com
Limit: Apple app access
An App Store reviewer complained that meaningful functionality is gated behind a paid multi-tenant account.Read 20 Oct 2026 · apps.apple.com
Limit: G2 complexity and cost
G2 reviewers repeatedly mention an overwhelming interface, documentation or guidance gaps, sluggish charts on large projects and rapidly rising storage or per-user costs.Read 20 Oct 2026 · g2.com
Limit: independent review pricing
Aicoolies says ingestion is usage-based and can make total cost difficult to estimate before a representative trial.Read 20 Oct 2026 · aicoolies.com
Limit: TrustRadius
TrustRadius reviewers report dashboard lag with many metrics and say custom-chart support and documentation are not straightforward.Read 20 Oct 2026 · trustradius.com
MCP extensibility
W&B skills and an MCP server let coding agents such as Claude Code read production data, run evaluations and iterate automatically.Read 20 Oct 2026 · site.wandb.ai
Monitoring
Built-in and custom signals capture and classify agent interactions, and alerts can be routed through Slack notifications and webhook automations.Read 30 Sep 2026 · site.wandb.ai
Notable release
Weave version 0.53.12 was released on 13 October 2026 with trace batching, tool-call ordering, newer OpenTelemetry exporters and TypeScript agent-schema updates.Read 20 Oct 2026 · github.com
OpenTelemetry API
Weave accepts OpenTelemetry spans at its OTLP endpoint, so an agent can send traces without installing the Weave SDK.Read 20 Oct 2026 · github.com
Playground
Playground tests prompts and custom models against production traces before deployment or during troubleshooting.Read 20 Oct 2026 · site.wandb.ai
Praise: G2 usability
G2 reviewers repeatedly praise ease of use, broad ML-framework integrations and an intuitive dashboard for comparing runs and collaborating.Read 20 Oct 2026 · g2.com
Praise: Gartner reviewer
The sole Gartner reviewer praised product and solutions-engineering support and described the platform as an AI development strategy with built-in practices.Read 20 Oct 2026 · gartner.com
Praise: independent review fit
Aicoolies rates Weave 82/100 and says it is especially suitable when tracing and evaluation should connect to an existing W&B workflow.Read 20 Oct 2026 · aicoolies.com
Praise: TrustRadius
TrustRadius reviewers praise metrics logging, hyperparameter sweeps, artifact tracking, reports and reproducibility.Read 20 Oct 2026 · trustradius.com
Rating on Apple App Store
4.1 out of 5 from 14 ratingsRead 20 Oct 2026 · apps.apple.com
Rating on G2
4.5 out of 5 from 83 reviewsRead 20 Oct 2026 · g2.com
Rating on Gartner Peer Insights
5.0 out of 5 from 1 reviewRead 20 Oct 2026 · gartner.com
Rating on TrustRadius
10 out of 10 from 2 reviewsRead 20 Oct 2026 · trustradius.com
SDKs and API
Weave provides Python and TypeScript libraries and supports sending OpenTelemetry traces to its OTLP endpoint.Read 30 Sep 2026 · docs.wandb.ai
SDKs and install
Official documentation provides Python and TypeScript libraries installed with pip install weave or npm install weave and requires a W&B account plus API key.Read 20 Oct 2026 · github.com
Self-hosting limit
The pricing comparison states that bring-your-own-bucket storage is not available with Weave.Read 20 Oct 2026 · site.wandb.ai
System requirements
The public Weave repository requires Python 3.10 or newer, a W&B account and a W&B API key.Read 20 Oct 2026 · github.com
Tracing
Weave organizes traces into sessions and turns and represents steps, tools, and sub-agents as first-class concepts.Read 30 Sep 2026 · site.wandb.ai
Usage limits
The pricing comparison lists 1 GB per month of Weave data ingestion on Free and 1.5 GB per month on Pro, with additional Pro ingestion priced at $0.10 per MB.Read 30 Sep 2026 · site.wandb.ai

Plans and prices

Free Free 1 GB/mo Weave data ingestion · 5 GB/mo storage · AI application evaluations, tracing, and scorers Read 30 Sep 2026 · site.wandb.ai
Free Free AI application evaluations · AI application tracing · AI application scorers · model experiment tracking · asset registry and lineage · community support · 5 GB/mo storage · 1 GB/mo Weave ingestion Read 20 Oct 2026 · site.wandb.ai
Pro $60 a month billed monthly Unlimited teams · team access controls · service accounts · priority email and chat support · CI/CD automations · Slack and email alerts · 100 GB/mo storage · 1.5 GB/mo Weave ingestion Read 20 Oct 2026 · site.wandb.ai
Enterprise Priced on request HIPAA-compliant option · private connectivity · customer-managed encryption keys · SSO · automated provisioning · custom roles · audit logs · enterprise support Read 20 Oct 2026 · site.wandb.ai

Every W&B Weave plan and price →

Strong points and limits

Strong points

  • Traces represent steps, tools, and sub-agents individually.
  • Alerts can route through Slack or webhooks.
  • Evaluations compare results and visualize regressions.
  • Free plan includes 1 GB monthly ingestion and 5 GB storage.
  • Supports Python, TypeScript, and OpenTelemetry traces.

Limits

  • Pro costs $60 a month.
  • Free ingestion is limited to 1 GB per month.
  • Additional Pro ingestion costs $0.10 per MB.

Who it suits

Weave suits teams monitoring and evaluating production AI agents, particularly those using supported frameworks and models. Its Free plan may fit smaller-scale use; Pro lists fewer than 50 employees and additional ingestion limits.

W&B Weave in depth

Read the full write-up

Overview

W&B Weave brings observability and continuous improvement to production AI agents, combining trace inspection with monitoring and evaluation. It suits teams building language-model applications that need to understand multi-step agent behavior and compare changes before deployment. Its broad toolkit is balanced by a modest monthly ingestion allowance on the paid entry plan.

Weights & Biases, founded in 2017 in San Francisco, was acquired by CoreWeave in May 2025. Weave also tracks token costs, prompts and retrieval activity, placing it across AI Agent Observability Tools, AI LLM Observability Tools, LLM Observability Tools, Model Monitoring Software and AI Agent Evaluation Tools.

Key features

Tracing and monitoring

Weave groups traces into sessions and turns, while treating steps, tools and sub-agents as distinct parts of an interaction. That structure is useful for following a multi-stage agent run instead of treating each model call as an isolated event. Built-in and custom signals classify interactions, and alerts can be routed through Slack notifications or webhook automations.

Cost tracking, token tracking, LLM tracing, prompt management and retrieval tracing make Weave relevant beyond agent workflows. Its Python and TypeScript libraries, plus an OTLP endpoint for OpenTelemetry traces, give teams several ways to send data into the platform.

Evaluation and model comparison

The evaluation framework pairs comparisons with visualizations to help teams spot regressions before deployment. In Playground, users can compare models and prompting techniques against production traces, connecting experiments to observed behavior rather than relying only on abstract test cases.

Pre-built safety scorers cover toxicity, bias, personally identifiable information and hallucinations. Quality scoring includes coherence, fluency and context relevance. This range supports both safety review and response-quality assessment, though the value depends on whether those score types match a team's evaluation criteria.

Integrations and security

Weave integrates with frameworks including LangChain and LlamaIndex and supports models from OpenAI, Meta, Amazon Bedrock and Anthropic. W&B states that its platform is certified to ISO/IEC 27001:2022, ISO/IEC 27017:2015 and ISO/IEC 27018:2019, and compliant with SOC 2 Type 2 and HIPAA standards. Data is encrypted in transit with TLS 1.2+ and at rest with AES 256. Enterprise options include single sign-on, audit logs, secure private connectivity and customer-managed encryption keys.

Pricing

Weave uses a freemium model, with a free plan, a 30-day trial and paid plans from $60 a month. The trial offers a low-commitment way to assess fit, while the free plan has ongoing usage caps.

PlanPrice and termsWhat it includesFreefree1 GB/mo Weave data ingestion, 5 GB/mo storage, AI application evaluations, tracing and scorers.Pro$60 a month, billed monthly; annual billing also available1.5 GB/mo ingestion, with additional ingestion at $0.10/MB. Starts with fewer than 50 employees; includes priority email and chat support.EnterpriseCustom pricing; invoiced annually upfrontCustomizable usage, enterprise support and security and compliance features.Free is a sensible starting point for small workloads or an initial evaluation, but its 1 GB monthly ingestion cap can constrain sustained production tracing. Pro raises that cap by only 0.5 GB per month, and overage at $0.10/MB makes ingestion volume an important cost variable. The Pro plan is aimed at organizations with fewer than 50 employees; larger or security-focused deployments may need Enterprise, where usage can be customized but pricing is negotiated and invoiced annually upfront.

Platforms

Weave supports API, Linux, macOS, self-hosted, web and Windows environments. That breadth, together with both self-hosted and hosted deployment options, gives teams flexibility in how they build and operate integrations.

Who it's for

Weave is a strong fit for teams operating production AI agents that need to trace tools and sub-agents, watch interaction signals, evaluate outputs and compare prompts or models against real traces. Its integrations and multiple SDK/API paths also suit teams already working with supported frameworks and model providers. It is less compelling for teams with high trace volume on a fixed budget: the Free and Pro ingestion allowances are limited, and Pro overage is charged per MB.

Pros and cons

Pros: Sessions, turns, steps, tools and sub-agents are all represented in the tracing model, giving multi-step agent activity a clear structure.

Pros: Monitoring, evaluations, production-trace comparisons and safety and quality scorers sit in one toolkit, supporting a workflow from observation to pre-deployment review.

Pros: Framework and provider integrations, Python and TypeScript libraries, and an OTLP endpoint offer multiple routes for connecting applications.

Cons: Free ingestion is capped at 1 GB per month, and Pro at 1.5 GB per month; additional Pro ingestion costs $0.10/MB.

Cons: Enterprise uses custom pricing and annual upfront invoicing, so it is a less straightforward choice for buyers who need a predictable published price.

Alternatives

PetalTrace is a free alternative with API, desktop, web and self-hosted platform support; consider it when free access is the priority. SigNoz is another freemium option, with a self-hosted community plan that puts infrastructure, storage, scaling, upgrades and backups under the user's management. Opik makes sense for teams that want its core observability and evaluation features as open-source code to download and run locally. OpenLIT is worth considering when unlimited self-hosted usage, users and projects are the deciding factors; its OSS plan includes community support.

Traccia offers a free Hobby plan with 50K events and seven days of retention, or an Observe plan at $99 a month with 500K events. VoltOps has a free Developer plan with one seat, one prompt, one project and 250 traces per month. Helicone is a freemium, open-source alternative. Maxim AI offers a free Developer plan for up to three seats, one workspace and 10k monthly logs, with three-day data retention.

Verdict

Choose W&B Weave if your team needs a combined view of agent traces, monitoring and evaluation, especially when reviewing multi-step behavior and testing model or prompt changes against production activity. Its strongest case is the combination of structured agent tracing and evaluation workflows. Look elsewhere if your expected ingestion is high and cost predictability is paramount: Pro's monthly allowance is limited, overages are metered, and Enterprise pricing is custom.

Verdict

Weave combines agent tracing, monitoring, evaluations, and model comparison with integrations for several frameworks and providers. The Free plan has defined ingestion and storage limits, while Pro starts at $60 a month.

Where Specifiction places it

Alternatives to W&B WeaveAll alternatives →