Connect Weights & Biases to Your AI Agent

Ask your AI assistant to work with Weights & Biases: compare experiments and investigate weave traces.

Try a Weights & Biases Task

Copy a prompt into your agent with the oo Skill installed. Add your page links and source notes before running it.

oo Skill Weights & Biases

Compare Training Runs

Compare configuration and metrics for two known runs.

/oo Weights & Biases: Compare W&B runs [run A ID] and [run B ID] in entity [entity] project [project], then retrieve their sampled histories to explain metric differences.

Investigate Trace Patterns

Discover trace fields before querying and counting records.

/oo Weights & Biases: Infer classic Weave trace schema in [entity/project], query traces matching [criteria], and count matching traces; summarize inspected examples separately from counts.

Start Working with Weights & Biases

Install OOMOL

Set up oo CLI and the oo Skill in your agent.
Install OOMOL →
Connect Weights & Biases

Connect using your Weights & Biases W&B API Key.
Connect Weights & Biases →
Give Your Agent a Task

Use /oo with the pages to read, the changes to make, and where to save the result.
Connection Guide →

Choose Your Agent

Codex Claude Code Cursor Hermes

An account with access to the Weights & Biases resources used in your task.

Weights & Biases Tools

Weights & Biases connects experiment metrics, artifact context, and classic Weave traces to AI investigation and reporting.

Compare Training Runs

Compare configuration and metrics for two known runs.

Investigate Trace Patterns

Discover trace fields before querying and counting records.

22 tools

Query Weave Traces

Query classic Weave traces with filters, ordering, selected columns, cost data, feedback, and detail-level controls.

wandb.query_weave_traces

Count Weave Traces

Count matching classic Weave traces and root traces without returning trace payloads.

wandb.count_weave_traces

Resolve Trace Roots

Resolve the root spans for a batch of classic Weave trace IDs.

wandb.resolve_trace_roots

Query Wandb

Run a read-only GraphQL query against W&B experiment data with bounded pagination.

wandb.query_wandb

Create Report

Create a W&B report with narrative text, plots, and configurable panels.

wandb.create_report

Log Analysis

Log analysis rows, charts, and scalar metrics to W&B as a new run.

wandb.log_analysis

List Entities

List W&B user and team entities accessible with the configured API key.

wandb.list_entities

List Projects

List W&B projects for one entity or for all entities accessible with the configured API key.

wandb.list_projects

List Automations

List W&B Automations that trigger on artifact, run-state, or run-metric events.

wandb.list_automations

List Integrations

List Slack and webhook integrations available as W&B Automation targets.

wandb.list_integrations

Infer Trace Schema

Sample classic Weave traces to discover field paths, types, and frequent values.

wandb.infer_trace_schema

Search Docs

Search the official W&B documentation for relevant guidance and examples.

wandb.search_docs

Get Run History

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

wandb.get_run_history

List Registries

List W&B model registries for an organization.

wandb.list_registries

List Registry Collections

List artifact collections within a W&B model registry.

wandb.list_registry_collections

List Artifact Versions

List versions of a W&B project artifact or registry collection.

wandb.list_artifact_versions

Get Artifact Details

Get metadata, lineage, and optionally files for a W&B artifact version.

wandb.get_artifact_details

Compare Artifact Versions

Compare metadata, aliases, tags, lineage, sizes, and optionally files for two W&B artifact versions.

wandb.compare_artifact_versions

Compare Runs

Compare configuration, summary metrics, metadata, and optionally metric history for two W&B runs.

wandb.compare_runs

Summarize Evaluation

Aggregate classic Weave evaluation results for a project or named evaluation.

wandb.summarize_evaluation

Diagnose Run

Inspect loss metrics and run state to diagnose the training health of a W&B run.

wandb.diagnose_run

Probe Project

Sample W&B runs to discover a project's metrics, configuration keys, tags, groups, and structure.

wandb.probe_project

Frequently Asked Questions

Does the hosted MCP support Multi-tenant Cloud?

The hosted endpoint supports Dedicated Cloud, not Multi-tenant Cloud; Self-Managed deployments need a separate public server.

Is run history exhaustive?

The history action retrieves sampled time-series metrics.

Can I save an analysis?

Create a report or log analysis data as a new W&B run with explicit entity and project.