Weights & Biases
The official implementation of the W&B Models and Weave MCP server.
What is Weights & Biases?
Weights & Biases is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to official implementation of the w&b models and weave mcp server.
The official implementation of the W&B Models and Weave MCP server.
This server falls under the Data Science & ML category on MCPgee, the world's largest MCP server directory with 33,000+ servers.
Features
- The official implementation of the W&B Models and Weave MCP
Use Cases
Maintainer
Works with
Installation
Manual Installation
npx wandbConfiguration
Configuration Details
claude_desktop_config.json
Performance
Response Metrics
Resource Usage
How to Set Up and Use Weights & Biases
The W&B MCP Server is the official Weights & Biases integration for AI assistants, providing 14 tools that let you query ML experiment runs, analyze LLM traces from Weave, manage model registry artifacts, generate W&B reports, and search W&B documentation — all through natural language. It supports both a hosted endpoint (https://mcp.withwandb.com/mcp) for instant use in Claude Code and a local self-hosted mode via uvx for Claude Desktop, making it accessible whether you need quick experiment lookups or deep artifact comparison workflows.
Prerequisites
- A Weights & Biases account and API key from https://wandb.ai/authorize
- For local installation: uv package manager installed
- Claude Code or Claude Desktop as your MCP client
- Active W&B runs, Weave traces, or model registry artifacts to query (the server reads existing W&B data)
Get your W&B API key
Log in to wandb.ai and visit https://wandb.ai/authorize to retrieve your personal API key. Keep this key secure — it provides full read/write access to your W&B account.
Option A: Add the hosted MCP server to Claude Code
For Claude Code, connect directly to the hosted W&B MCP endpoint over HTTP. This requires no local installation.
claude mcp add --transport http wandb https://mcp.withwandb.com/mcp \
--scope user --header "Authorization: Bearer YOUR_WANDB_API_KEY"Option B: Install locally for Claude Desktop
For Claude Desktop, add the server configuration using uvx to run the package directly from the GitHub repository.
{
"mcpServers": {
"wandb": {
"command": "uvx",
"args": ["--from", "git+https://github.com/wandb/wandb-mcp-server", "wandb_mcp_server"],
"env": {
"WANDB_API_KEY": "your-api-key-here"
}
}
}
}Configure optional environment variables
For on-premises or dedicated W&B deployments, set WANDB_BASE_URL. Adjust MAX_RESPONSE_TOKENS if responses are being truncated. Set MCP_SERVER_LOG_LEVEL for debugging.
export WANDB_BASE_URL=https://your-org.wandb.io # only for on-prem
export MAX_RESPONSE_TOKENS=30000
export MCP_SERVER_LOG_LEVEL=INFOVerify the connection
Ask Claude to list your W&B projects or query your most recent experiment run to confirm the server is connected and authenticated.
Weights & Biases Examples
Client configuration
Claude Desktop configuration using uvx to run the W&B MCP server locally with API key authentication.
{
"mcpServers": {
"wandb": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/wandb/wandb-mcp-server",
"wandb_mcp_server"
],
"env": {
"WANDB_API_KEY": "your-api-key-here",
"WANDB_BASE_URL": "https://api.wandb.ai",
"MAX_RESPONSE_TOKENS": "30000"
}
}
}
}Prompts to try
Use these prompts to explore your ML experiments, traces, and model artifacts through Claude.
- "Show me the top 5 runs from my MNIST project sorted by validation accuracy"
- "Analyze the latency distribution of my GPT-4 Weave traces from last week"
- "Compare the two latest versions of the 'image-classifier' artifact in my model registry"
- "Create a W&B report summarizing the hyperparameter sweep results from my latest experiment"
- "Search W&B docs for how to use custom metrics with Weave evaluations"Troubleshooting Weights & Biases
Authentication fails with 401 Unauthorized
Verify your WANDB_API_KEY is correct by testing it with: curl -H 'Authorization: Bearer YOUR_KEY' https://api.wandb.ai/graphql. If using a dedicated instance, ensure WANDB_BASE_URL points to your organization's W&B URL, not the public api.wandb.ai.
uvx command not found when starting the local server
Install uv using: curl -LsSf https://astral.sh/uv/install.sh | sh (macOS/Linux) or 'winget install astral-sh.uv' (Windows). After installation, use the full path to uvx in your config (find it with 'which uvx').
Responses are truncated or missing data from large experiments
Increase the MAX_RESPONSE_TOKENS environment variable (default is 30000). For very large run histories, ask Claude to filter by specific metrics or date ranges to reduce response size.
Frequently Asked Questions about Weights & Biases
What is Weights & Biases?
Weights & Biases is a Model Context Protocol (MCP) server that official implementation of the w&b models and weave mcp server. It connects AI assistants to external tools and data sources through a standardized interface.
How do I install Weights & Biases?
Follow the installation instructions on the Weights & Biases GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.
Which AI clients work with Weights & Biases?
Weights & Biases works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.
Is Weights & Biases free to use?
Yes, Weights & Biases is open source and available under the MIT license. You can use it freely in both personal and commercial projects.
Weights & Biases Alternatives — Similar Data Science & ML Servers
Looking for alternatives to Weights & Biases? Here are other popular data science & ml servers you can use with Claude, Cursor, and VS Code.
Ultrarag
★ 5.6kA Low-Code MCP Framework for Building Complex and Innovative RAG Pipelines
RocketRide
★ 3.1k📇 🏠 - MCP server that exposes RocketRide AI pipelines as t
Aix Db
★ 2.1kAix-DB 基于 LangChain/LangGraph 框架，结合 MCP Skills 多智能体协作架构，实现自然语言到数据洞察的端到端转换。
NeMo Data Designer
★ 1.9k🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data.
PaperBanana
★ 1.7kOpen source implementation and extension of Google Research’s PaperBanana for automated academic figures, diagrams, and research visuals, expanded to new domains like slide generation.
MiniMax
★ 1.5kBridges MiniMax AI capabilities to the Model Context Protocol, enabling AI agents to perform image understanding, text-to-image generation, and speech synthesis. It provides a standardized interface for accessing MiniMax's core tools via JSON-RPC.
Browse More Data Science & ML MCP Servers
Explore all data science & ml servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.
Set Up Weights & Biases in Your Editor
Choose your AI client for step-by-step setup instructions.
Quick Config Preview
Add this to your claude_desktop_config.json or .cursor/mcp.json
Ready to use Weights & Biases?
Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.