MCP server for mechanistic interpretability research, enabling agents to drive probe-causality and SAE-feature experiments via 8 typed tools on user's own compute (Colab).
Mechanistic interpretability MCP server wrapping chuk-lazarus, enabling model loading, activation extraction, probe training, steering, and ablation via MCP tools.
An MCP server that records agent execution metrics and exposes a Context Window Explorer to visualize exactly what entered the model's context window across sessions, tokens, and tool calls.