Scout

What it does

Scout inspects one result of a segmentation, a registration or a synthesis the way a person would at a viewer: it looks at slices, zooms, switches overlays, asks KonfAI Eval for measurements on a region, and writes a report of localised defects with the views that show them. It is not a scorer: KonfAI Eval scores, Scout reads those scores and asks for more on a region. Every location is in millimetres. The report keeps three kinds of statement apart, a measured property, a suspected error and a confirmed error, and a confirmation must cite a measurement with a reference: something seen but not measured stays a suspicion.

Install and run

pip install konfai-scout[claude-code]      # Claude under your Claude Code login, no API key
pip install konfai-scout[codex]            # GPT under your ChatGPT login (the Codex CLI), no API key
pip install konfai-scout[anthropic]        # Claude through the Anthropic API
pip install konfai-scout[openai]           # GPT, or any OpenAI-compatible server (vLLM, Ollama, LM Studio)

Images leave the machine only when you allow it (allow_cloud_images); a local server is local. The MCP server runs Scout through run_scout_inspection and Studio shows the report in its SCOUT tab: verdict, observations, measurements and the cited views. Details: Scout on this site and the konfai-scout folder (its docs/ hold the report schema and the study protocol).