W&B

Queries and analyzes W&B data via natural language using the MCP, enabling actions like analyzing runs, traces, and reports.
  • typescript

7

GitHub Stars

typescript

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

You can query and analyze your Weights & Biases data through natural language using the Model Context Protocol. This MCP server lets you ask questions about runs, traces, reports, and more, delivering fast insights directly to your preferred MCP-enabled client.

How to use

Connect to the MCP server from your client of choice and start asking questions about your W&B data. You can target specific projects and entities, refine questions to get precise results, and rely on the tools available to fetch runs, traces, and reports.

How to install

Choose your deployment path based on your needs. You have a hosted option for zero-configuration access and a local option for full control.

Option 1 Hosted Server
No installation needed. Use the hosted MCP at:
https://mcp.withwandb.com/mcp

Additional configuration and usage notes

HTTP connection (Hosted) and local STDIO connection (Self-hosted) are supported. If you run locally, you can expose an HTTP endpoint for client access or use STDIO with desktop clients.

HTTP (Hosted) example
URL: https://mcp.withwandb.com/mcp
No local server to run

STDIO (Local) example
command: uvx
args: ["wandb-mcp-server","--transport","http","--port","8080"]

Security and keys

Keep your W&B API key secure. When using HTTP-based or local transports, pass credentials as required by your client configuration. Do not expose keys in public or shared environments.

Troubleshooting

If you cannot reach the MCP server, verify the URL, ensure your API key is valid, and check that the transport and port settings match your client configuration.

Available tools

query_wandb_tool

Query W&B runs, metrics, and experiments to retrieve specific data like top losses, best runs, and comparisons.

query_weave_traces_tool

Analyze LLM traces and evaluations to evaluate latency, quality, or other metrics.

count_weave_traces_tool

Count traces and gather storage metrics to understand data volume and usage.

create_wandb_report_tool

Programmatically create W&B reports to compile results and visualizations.

query_wandb_entity_projects

List projects owned by an entity to scope your queries.

query_wandb_support_bot

Get help from W&B documentation and guidance for MCP usage.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational