Graph editor
Devices, ports, and connections in one canvas. Drag-to-connect with port-type validation.
MCP-native lab context
Your lab as a graph. Ready for your AI.
DCS models your lab setup as a structured graph of devices, connections, and parameters. Photos, ELN imports, and equipment manuals become context for Claude, Codex, and Gemini — so answers reflect your actual setup, not generic knowledge.
Open .dcs.yaml format · We're open to proposals — academic and industry
Works with Claude Code Codex Gemini VSCode 1.85+
Used at research groups including
TU DarmstadtResearch labResearch labResearch labResearch lab
01 · Model
Model your lab as a graph
Every device is a node. Every cable, tube, fiber, or data link is an edge. Port types catch incompatible connections before you wire them — no more 'why isn't the signal arriving' at midnight.
- Validated port compatibility (electrical, optical, fluid, gas, data, thermal, mechanical)
- Live diff of the rig — see what changed between yesterday and today
- .dcs.yaml is git-friendly and human-readable
version: 1
lab: chromatography-bench
devices:
pump:
type: hplc-pump
model: agilent-1290
ports:
out: { signal: fluid }
column:
type: chromatography-column
model: zorbax-eclipse-c18
parameters:
length: 100mm
particle_size: 1.8um
detector:
type: uv-detector
model: agilent-g7117
ports:
in: { signal: fluid }
data_out: { signal: data }
connections:
- from: pump.out
to: column.in
signal: fluid
- from: column.out
to: detector.in
signal: fluid
- from: detector.data_out
to: workstation.data_in
signal: data
02 · Capture
Photos, manuals, and history — all attached
Snap a photo on the bench, the mobile app finds the device by QR. PDF manuals get indexed for semantic search. Every parameter change records itself with a timestamp.
- PWA on any phone — install from the browser, works offline
- Manuals: drop the PDF, ask questions, get cited answers
- Per-parameter history without git knowledge
Screenshot pending — DCS mobile · scan & attach
03 · Plug in
Your AI tools get the whole lab as context
Claude Code, Codex, and Gemini speak MCP. DCS publishes your graph as MCP tools — answers are grounded in your actual setup, not generic knowledge.
- Zero-config: install the extension, MCP is live
- Tools: read graph, read parameters, search manuals, query history
- Local-first — your data never leaves the workspace
Screenshot pending — Claude Code · MCP · DCS
Everything the rig needs, none of the integration tax
One workspace covers the modeling, capture, indexing, and AI integration that today's lab pieces together by hand.
Mobile photos
Scan a QR sticker, photograph the rack, auto-attach to the right device — all in seconds.
PDF indexing
Equipment manuals indexed for semantic search. Ask "what is the max pressure for column X" — get the page.
MCP tools
Claude Code, Codex, and Gemini receive your full lab as context — no copy-paste, no hallucinations.
Bring your own ELN
Import from eLabFTW, openBIS, or SciCat in seconds via the open ELN RO-Crate standard. DCS complements your ELN — it does not replace it.
DSL format
.dcs.yaml is readable, validated, and git-friendly. Diff a change to the rig like you diff code.
"We had three different schematics, four spreadsheets, and a Slack thread. DCS made one canvas the lab actually agrees on."
— Pilot research group, TU DarmstadtThree goals shaping every release
DCS is a wedge into a bigger ambition: an open ecosystem where lab work is fast, devices know themselves, and live data flows naturally into models researchers can trust.
01
An open standard for the lab
Propose .dcs.yaml — and the surrounding ecosystem of integrations and tooling — as the open layer that makes lab work fast and friction-free. Open by default, never vendor-locked.
02
The most complete device catalog
A community-curated catalog of laboratory devices — ports, parameters, manuals, constraints, real-world quirks. The bigger it grows, the smarter every AI assistant becomes in the lab.
03
Digital twins, live
Bridge static device models with live monitoring and analytics. The lab graph isn't a diagram — it becomes a digital twin that knows what is running, what is measured, and what changed last week.
Questions we hear from research groups
How is DCS licensed and priced?
DCS is in early access. Pricing is scoped per engagement — talk to us and we will shape a pilot that fits your lab. The .dcs.yaml data format is an open standard and we intend to open-source it, so your lab data is never locked in.
We already use eLabFTW / openBIS / SciCat. Do we have to switch?
No. DCS complements your ELN — it does not replace it. We import RO-Crate (.eln) exports in 30 seconds via the open ELN Consortium standard. Bidirectional sync is on the roadmap; if you have a specific integration in mind, we are happy to scope it together.
Why VSCode and not a desktop app?
Researchers already install VSCode for Python/R/Jupyter. Zero-friction install plus native MCP server integration with Claude Code. A standalone web dashboard is on the roadmap.
Which AI tools work with DCS?
Anything that speaks Model Context Protocol — Claude Code, Codex, Gemini, custom MCP clients. Your lab context shows up automatically once the extension is installed.
What about 21 CFR Part 11 / GLP compliance?
The architecture already records an immutable audit trail of every change. Formal certification is on the post-pitch roadmap — not a blocker for academic and early-biotech labs. If you need a specific compliance path, we are open to scoping it as part of a pilot.
Is my lab data stored in the cloud?
No. The .dcs.yaml file lives in your workspace, the local database is in .dcs/dcs.db, and the MCP server runs in-process inside VSCode. Optional sync is opt-in. On-prem deployments for security-sensitive labs are available — talk to us.
Bring DCS to your bench
Tell us about your lab. We shape pilots, integrations, and pricing around what your setup actually needs — academic groups and industry teams welcome.