Human / Moment

An evidence-linked, interactive 3D journey through the anatomy and timing of a human sneeze.

Launch the public atlas · No account, API key, or runtime AI required.

Run locally

npm install
npm run dev

Production verification:

npm test
npm run lint
npm run typecheck
npm run build
npm run test:e2e

ChatGPT app integration (no API key)

The repository also contains one deterministic, read-only MCP tool, open_sneeze_atlas. It returns the same canonical stage, layer, timing, evidence, and source data used by the website. It does not call an OpenAI model or any external medical service. The tool is linked to a versioned, self-contained ui:// Apps SDK resource that renders a compact scene card in ChatGPT without embedding the public site in a nested iframe. Its primary action opens the exact stage, camera focus, and visible-layer combination in the full 3D atlas through a deterministic deep link.

Public MCP endpoint: https://human-moment-mcp.ozanpochette.workers.dev/mcp

cp .env.example .env
npm run build
npm run start:mcp

The local endpoints are http://127.0.0.1:8787/healthz and http://127.0.0.1:8787/mcp. Test the latter with MCP Inspector. To connect from ChatGPT developer mode, use the public HTTPS endpoint above. The same stateless contract is deployed as a Cloudflare Worker; npm run dev:mcp:worker exercises that runtime locally and npm run deploy:mcp publishes it. A static host can serve the website, but it cannot serve the MCP endpoint.

For a container deployment, set MCP_HOST=0.0.0.0, the platform-provided PORT, PUBLIC_APP_URL, and ALLOWED_HOSTS=<your-mcp-host>. None are secrets.

What makes it different

• Real NIH / Human Reference Atlas geometry instead of generated anatomy.
• A deterministic six-stage physiology timeline with visible evidence confidence.
• Cinematic guided playback and free 3D exploration in one experience.
• No API key or account required for the public website or MCP server.

OpenAI Build Week

Human / Moment is being built with Codex and GPT-5.6. Codex is used for implementation, asset-pipeline tooling, test automation, and source-grounded content refinement. The anatomy and medical claims remain static and inspectable. The thin, read-only ChatGPT Apps SDK integration controls canonical scene parameters rather than generating medical content.

See the implementation design and asset attributions.