Keep pulling the thread on Weights & Biases.
Weights & Biases now offers a completely hosted version of its Model Context Protocol (MCP).
Weights & Biases has released an update to its Model Context Protocol (MCP), which has been available as a public repository for the last couple of months.
The hosted Weights & Biases Model Context Protocol (MCP) is available for all deployment types, including SaaS, dedicated, and on-prem.
The Model Context Protocol (MCP) is a standardized format that enables agents to interact with tools, specifically allowing them to access Weights & Biases data, experiments, and traces.
The Weights & Biases Model Context Protocol (MCP) integrates with IDEs such as Cursor and VS Code.
The Weights & Biases Model Context Protocol (MCP) supports integrations with coding agents including Claude Code, Codex, and Gemini CLI.
The Weights & Biases Model Context Protocol (MCP) integrates with CAT agents such as Blue Shell and Cloud Desktop.
The Weights & Biases MCP includes a 'Discovery' toolset that enables agents to autonomously identify teams, projects, data within projects, and experiment schemas.
The Weights & Biases Model Context Protocol (MCP) provides tools for agents to query run information, compare runs, and diagnose runs.
Agents can use the Weights & Biases Model Context Protocol (MCP) to query traces, count traces, and summarize evaluations.
The Weights & Biases MCP includes tools that allow agents to create reports and perform analysis before the report is generated.
Weights & Biases offers a serverless GRPO training endpoint for model optimization.