Weights & Biases

By thumbsupordown

The official Weights & Biases MCP server connects AI agents to your W&B Models and Weave data. Agents can query experiment runs, metrics, and sweeps, analyze and count Weave LLM traces, browse the model registry and artifacts, search the docs, and create W&B reports — grounding assistants in your experiment-tracking data.

Tools

compare_artifact_versions_tool

Diff two versions of a W&B artifact and show changes.

count_weave_traces_tool

Count Weave traces and retrieve storage size metrics.

create_wandb_report_tool

Create a W&B report with markdown, charts, and panels.

get_artifact_details_tool

Get the full details of a specific W&B artifact version.

get_run_history_tool

Retrieve sampled time-series metric data for a W&B run.

infer_trace_schema_tool

Discover field names, types, and sample values in Weave traces.

list_artifact_versions_tool

List every version of a given W&B artifact collection.

list_registries_tool

List the model registries within a W&B organization.

list_registry_collections_tool

List the artifact collections within a W&B registry.

list_wandb_automations_tool

List the configured W&B Automations for a project.

list_wandb_integrations_tool

List registered integrations used by W&B automations.

log_analysis_to_wandb

Log analysis results back to W&B as a new tracked run.

query_wandb_entity_projects

List all of the projects belonging to a W&B entity.

query_wandb_tool

Query W&B runs, metrics, and experiment tracking data.

query_weave_traces_tool

Analyze LLM traces from Weave with detail-level control.

search_wandb_docs_tool

Search the official Weights & Biases documentation.

Want to create your own MCP server?

Create an account to add and manage your own MCP servers.