Built around
your documents.
Every current Droblet client runs a Private Cloud deployment we built for them: their document types, their systems, their customs flows, their output formats. If your operation does not fit a subscription, this is how we work with you.
Yours
Deployment, data and tenant
Weeks
To the first live document, not months
Founders
On every call
What custom means here
The engine is the same one every client uses: schema-driven extraction with confidence and provenance, case checks, declaration drafting with gaps, tariff, agents, MCP, and the delivery layer with its five formats. What we build for you sits on top of it: the schemas for your documents, the connector to your system, the output template your loader expects, the filing rail you need, the checks your team runs every morning.
That is why a custom build takes weeks and not months. Nothing is written from scratch. A new document type is a schema, usually inferred from your samples in an afternoon. A new output is a template. A new system is a connector on a delivery layer that already handles retries, idempotency and host-key pinning. A new scheduled check is a trigger.
You get your own deployment, in your cloud account or ours, with your data in it and nobody else's. We run it with you, and both founders are on the calls.
What we build
The pieces a Private Cloud project is usually made of.
Document schemas
Your invoices, certificates, orders, claims or records, described once and read the same way every time. Inferred from samples, then tuned with your team.
Connectors
Into SAP, Salesforce, Shopify, PennyLane, ELO, your TMS, your WMS, or a system nobody outside your company has heard of.
Output formats
JSON, XML, CSV, fixed-width or EDIFACT, in the exact layout your loader or trading partner expects. Templates you can edit afterwards.
Filing rails
HMRC CDS and EU ICS2 today. French ELO. Another authority or port system if your flows need it.
Checks and agents
The questions your team asks every day, run on a schedule or on an event, reported back without anyone asking.
MCP for your team
The Droblet MCP server on your host, so your people can work from Claude or ChatGPT with their own permissions and no way to file without approval.
How a project runs
Short, with something live early.
Discovery, one week
We take your three hardest documents and your target system, and come back with the fields, the gaps and the output in your format.
Build, two to four weeks
Schemas, connectors, templates and checks. You see progress on your own documents every week.
Pilot on live work
Your team runs real cases alongside the old way. We measure what gets flagged and tune the thresholds.
Run
Your deployment, your data, support from the people who built it. New document types and connectors added as you need them.
Common questions
It runs in your cloud account or in ours, dedicated to you. Your data never leaves your tenant and no other client shares it.
The shared platform runs the same engine, so a Private Cloud client can move to it when it launches, schemas and templates included.
No. The engine reads any repeatable document. We have built for finance teams reading supplier invoices and for operations teams reading certificates; customs is where it started.
A fixed project fee for discovery and build, then a monthly fee for the deployment and support. We quote after discovery, when we have seen the documents.
Start with your three hardest documents
One week of discovery, on your paperwork, with a founder. You see the result before you commit.