Skip to content

Data integration for AI coding

Data integration for AI coding: ship data-heavy apps with your agent

Your agent can write the app. It can't guess your data.

Developers, agencies and in-house teams use dcupl to build catalogs, customer portals and order entry on ERP and product data. Your coding agent builds on the real model of that data, and the frontend can start before the backend is finished.

npm install -g @dcupl/cli

The CLI is free on npm and runs locally. The Console has a free tier.

In production at industrial companies since 2022, from steel to eyewear.

  • voestalpine
  • Mehler Elektrotechnik
  • STRASSER Steine
  • LiSEC
  • Silhouette
  • Lumeso
The problem

Agents don't crash on bad data. They succeed on the wrong data.

  • Guessing

    The agent invents a field name, gets an empty array back and reports success.

  • Control

    Changes land faster than anyone can review them. Nothing says what is pending.

  • Time

    And the rollout deadline still looms next month.

Nothing crashed. Nothing is right.

Why we build dcupl

The hard part of an app was never the code. It's the data.

Agents made code cheap. So the data path now decides whether an app is right. We believe every team building with agents deserves data its agent can read, check and trust before anything ships.

How we do it

Give the agent the data, not a guess.

  1. Read the real model first

    The agent asks for actual fields and types before it writes a filter. Wrong names fail loudly, not silently.

  2. Answer in JSON

    With --json each command returns structured output. The agent knows what happened, and so do you.

  3. Nothing live without a build

    Changes are staged, workflows run on a dev runner first. A person approves what goes live.

The difference in one session. Same prompt, same data.

You, to both agents: import new-prices.csv and list products on sale under €50

Agent without dcuplguesses
agent

Merging new-prices.csv into products.json … done.

agent
products.filter(p => p.on_sale && p.price_eur < 50)[]
agent

Import complete. No products are on sale under €50.

The same guess, sent to the dcupl CLI

dcupl app query execute --model Product \ --query '[{"attribute":"on_sale","operator":"eq","value":true},{"attribute":"price_eur","operator":"lt","value":50}]' \ --json{"success":false,"error":"Attribute \"on_sale\" not found on model \"Product\". Did you mean \"onSale\"?","code":"ATTRIBUTE_NOT_FOUND"}

Exit code 1. The guess stops here, with the right name in the error.

Reported as done. The answer is wrong. The prices in products.json are already overwritten, and nothing failed, so nobody looks.

Agent with the dcupl CLIchecks

1: Read the model

dcupl app models get --model Product --json{"key":"Product","properties":[…,{"key":"price","type":"float"},{"key":"onSale","type":"boolean"}]}

2: Stage the import

dcupl app data upsert --file new-prices.csv \ --model Product --key-property id --json{"upserted":5,"model":"Product","built":false,…}
agent

5 price changes staged, not live. Build?

you

yes

3: Build after your yes

dcupl app build --json{"appId":"shop","built":true}

4: Query the built data

dcupl app query execute --model Product \ --query '[{"attribute":"onSale","operator":"eq","value":true},{"attribute":"price","operator":"lt","value":50}]' \ --sort price --projection name,price --json[{"key":"SKU-2001","name":"Rocker switch, white","price":6.9},{"key":"SKU-3002","name":"LED downlight, 7 W","price":12.4},…]
Show all 9 rows{"key":"SKU-2003","name":"Two-way switch, anthracite","price":14.2},{"key":"SKU-4002","name":"Smoke detector, 10-year","price":17.9},{"key":"SKU-1003","name":"Flat cable 3x1.5, 25 m","price":24.5},{"key":"SKU-3003","name":"LED strip 5 m, 4000 K","price":34.9},{"key":"SKU-4001","name":"Motion sensor, ceiling","price":39.5},{"key":"SKU-1001","name":"NYM-J cable 3x1.5, 50 m","price":39.9},{"key":"SKU-2002","name":"LED dimmer switch, white","price":44}]
agent

9 products on sale under €50, cheapest first.

The right 9 products, from the built data. The 5 new prices went live with the build, not before.

The commands and the CLI's replies were captured on twelve demo products; … marks a cut. The agent's lines and the session without dcupl are illustrative.

Why it works
What you get

One data layer, from the export to the app and the agent

  • 01

    CLI and skill

    How your agent drives dcupl. Works with any agent, Claude Code gets a skill.

  • 02

    Workflows

    Load, join and clean exports. Visible in the Console, tested before production.

  • 03

    Model

    Typed, joined across sources, with quality rules. One source of truth.

  • 04

    SDK, REST API, MCP

    The same model for your frontend, your backend and your client's agents, where MCP is turned on.

Exports in. A workflow from the dcupl Console, running.

Workflow · product-syncdcupl workflow test --file sync.workflow-v3.json --runner dev-runner
Run status: Done
  1. Sync request
    trigger-request
    fired
  2. PIM export
    azure-files
    done
  3. ERP prices
    request
    done
  4. Join and clean
    script
    done
  5. Product data
    dcupl-files
    done
  6. Respond
    response-status
    done
trace · run on dev runner
  1. trigger-request fired
  2. azure-files running
  3. azure-files done
  4. request running
  5. request done
  6. script running
  7. script done
  8. dcupl-files running
  9. dcupl-files done
  10. response-status running
  11. response-status done

A model out. Typed and queryable. Try it on your own CSV.

Try it on your own CSV

dcupl infers the model, builds the filters and lists the values that don't fit. Nothing is uploaded.

demo-export.csvDemo data13 rows · 7 columns
Model: 7 properties, key id

4 potential errors 1 note in 13 rows

Checked: types, empty cells, keys, row shape

  • price1 value lost part of its value: 24,50 was read as 24.
  • idSKU-2002 appears twice. The last row is kept.
  • stock1 value is not a whole number: n/a. It loads as empty.
  • onSale1 value is not true or false: yes. It loads as empty.
  • ean13 values start with 0, so the column stays text.
No filters. Pick a facet.
Rows of the file, filtered by the selected facets
id, 1 potential error, stringkeyname, stringcategory, stringprice, 1 potential error, floatonSale, 1 potential error, booleanstock, 1 potential error, intean, 1 note, string
SKU-1001 NYM-J cable 3x1.5, 50 m Cable 42.90 false 40 0401234500011
SKU-1002 NYM-J cable 5x2.5, 50 m Cable 118.00 false 12 0401234500028
SKU-1003 Flat cable 3x1.5, 25 m Cable 24.00 true 85 0401234500035
SKU-2001 Rocker switch, white Switch 6.90 true 250 0401234500042
SKU-2002 LED dimmer switch, white Switch 46.00 true 31 0401234500059
SKU-2003 Two-way switch, anthracite Switch 14.20 true 88 0401234500066
SKU-3001 LED panel 600x600, 36 W Lighting 89.00 false 9 0401234500073
SKU-3002 LED downlight, 7 W Lighting 12.40 true – 0401234500080
1–8 of 12 rows · min 6.90 · avg 50.98 · max 164.00

The agent workflow is new. The engine is not.

In production for industrial companies since 2022.

  • voestalpine
    500,000+
    orders, 5+ million documents
  • Mehler Elektrotechnik
    11,000+
    items in the product search
  • STRASSER Steine
    30,000+
    sinks, hobs and extractor fans to configure a worktop
  • LiSEC
    myLiSEC
    customer portal

All success stories

Latest insights

All insights →
Three rows, each with a change, a staged state and a production state: the CLI with built false and dcupl app build, a workflow variant and promote to base, and a REST API instance with status updating.

NewInsight · Data governance · November 5, 2026

Stage, build, approve: why nothing in dcupl happens by surprise

The CLI stages before it builds. A workflow variant runs on a dev runner before it is promoted. An API instance serves the old data until the update is done. One rule, three places, and why it matters when agents write the changes.

Give your agent the real data.

npm install -g @dcupl/cli