About Weave
W&B Weave is a toolkit for developing AI-powered applications, built by Weights & Biases. It provides observability and evaluation capabilities for agents and LLM applications, allowing users to trace and collect metrics, manually instrument LLM calls, and evaluate responses using LLM judges and custom scorers.
The platform supports tracing agents built with popular SDKs and harnesses, and offers built-in integrations for agent SDKs. It also provides an OTLP endpoint for sending OpenTelemetry spans, and Python and TypeScript libraries for installation. It is available in CoreWeave Forge accounts.
Key features
- Trace and collect metrics about agents built with popular SDKs and harnesses using Weave OTel-compatible SDK
- Manually instrument application's LLM calls and arbitrary functions to trace, version, and collect feedback
- Evaluate agent's or application's responses using LLM judges and custom scorers
- Built-in integrations for agent SDKs or harnesses to start tracing sessions, turns, LLM calls, and tool calls
- Send OpenTelemetry spans to the Agents view via OTLP endpoint without Weave SDK installation
- Python and TypeScript libraries available
Good fit for
- Trace an agent
- Evaluate a RAG application
- Build an evaluation pipeline using Weave scorers
Weave: questions and answers
- What is Weave used for?
- Weave is an observability and evaluation platform for tracking, testing, and improving language model applications, with support for tracing, evaluation, and integration with popular SDKs. It is a good fit for trace an agent, evaluate a RAG application and build an evaluation pipeline using Weave scorers.
- Is Weave open source?
- Yes. Weave is open source under the Apache-2.0 licence. Its source code is on GitHub at wandb/weave and is written mainly in Python.
- Is Weave free?
- Yes. Weave is open source, so the software itself is free to use. A managed cloud version is also available.
- Can I self-host Weave?
- Yes. Weave can be self-hosted on your own server or infrastructure.
- What is Weave an alternative to?
- Weave is an open-source alternative to Vellum. Other open-source alternatives to Vellum include Langfuse, Arize Phoenix and Dify.
- Is Weave actively maintained?
- Yes. The most recent commit to Weave was on 10 November 2026, and the latest release is v0.54.1, published on 5 November 2026. The project has 1.1k stars on GitHub, with 6 added in the last 30 days.
Open-source alternatives to Weave
See all
Langfuse
AI Infrastructure
Trace, evaluate, and improve AI agents with one open platform. Use production data to unde
OSSvs Weights & Biases★ 37k
Arize Phoenix
AI Infrastructure
AI Observability & Evaluation
OSSvs Weights & Biases★ 12k
Dify
AI Infrastructure
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collabo
OSSvs Gumloop★ 161k
Promptfoo
AI Infrastructure
Test your prompts, agents, and RAGs. Red teaming/pentesting/vulnerability scanning for AI.
MITvs Braintrust★ 27k
RAGFlow
AI Infrastructure
RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cu
Apache-2.0vs Vectara★ 94k
Langflow
AI Infrastructure
Langflow is a powerful tool for building and deploying AI-powered agents and workflows.
MITvs Gumloop★ 157k

Vellum