DCS Pitch

Device Context System — MVP demo

The problem

Research labs operate complex setups built over years. The researcher who configured everything leaves. Their successor spends weeks reconstructing what was known. A single hidden controller parameter can cause days of downtime.

The insight

The gap is not in documenting experiments — it's in preserving the operational knowledge required to run the setup: parameters, dependencies, physical state, institutional memory.

What DCS does

1. Models the lab setup as a structured graph (.dcs.yaml files)
2. Preserves the mental model — parameters, history, photos, manuals
3. Surfaces this as structured context to AI tools the lab already uses

Demo: 4 minutes

1. Graph — LC-MS setup loads instantly. Devices, connections, parameters.
2. Context — Click ESI Source → history shows "voltage changed 3500→4200V 2 months ago".
3. AI — Ask Claude Code "Why is pressure dropping after the MFC?" → grounded answer citing actual parameters + manual page 47.
4. Photo — Scan QR on phone → attach photo of capillary deposit → badge updates live in VSCode.

Market

~8,000–18,000 experimental research groups in Germany. Global: ~100,000. Adjacent markets: ELN, LIMS, lab automation. Entry price: €500–2,000 per group per year.

Status

MVP: VSCode extension + local server + mobile PWA. Zero external API required. Works with Claude Code subscription. No user setup beyond installing the extension.