OneCloud is an applied-AI company. We design and ship the AI that works on your own data — strategy, retrieval, agents, automation and modernization — and we run two platforms underneath it, so what we prototype is what you operate in production.
Fixed fee, from $5,000 · the audit credits toward the build
A ladder, not a lock-in. Climb only as far as the value justifies.
Ask your company anything. Get answers with receipts.
Legacy, understood. Then modernized.
sample report · not a real scan
Most AI vendors sell you either advice or software. We do both, and the advice is delivered on software we own — so a pilot never has to be rebuilt to become production.
Strategy, retrieval, agents, automation, governance and enablement — delivered as fixed-fee, time-boxed engagements with a named team. No open-ended retainers, no science projects.
The enterprise RAG and agent platform: cited answers on your own documents, permissions enforced inside the index, 24+ connectors, visual workflow automation and an MCP endpoint.
AI-native legacy modernization in four phases: understand the system, plan the migration, protect behavior with generated tests, then transform it — with evidence on every claim.
Eight disciplines, one team. Take the whole programme or the single piece you're missing — each capability below is something we have built, shipped and operated, not a slide.
Start where you are. The audit de-risks the build, and its fee is credited toward whichever build follows — so the first step never costs you twice.
Know exactly where AI pays off at your company — before you spend a dollar building it.
For leaders who know AI matters but need a defensible plan.
A production system on your own documents — cited answers, access control, live in weeks.
For teams with the documents and a clear use case, who want it in production.
Agents that do the work — from cited answers to real actions, all the way to the factory floor.
For organizations that want AI to act, not just answer.
Larger programmes and modernization assessments, priced per system. Every deliverable is produced on the platform, so you keep the workspace when the engagement ends.
A read-only archaeology of a target's codebase in days: score, risks, key-person exposure and a modernization estimate for the deal team.
Full archaeology plus a target profile, dispositions and a costed strangler-fig sequence with scenarios, reviewed with your architects.
The assessment, a generated safety net of characterization tests and the first verified change sets, packaged as a program plan.
We run the platform, the evals, the model budget and the on-call — a standing AI team for the years after the build, priced per quarter.
All prices in USD · delivered remotely worldwide · the audit fee is credited toward a subsequent build
Two products we build, ship and operate. Consulting engagements land on them, which is why a six-week pilot doesn't need a rewrite to become the system you run for the next five years.
The enterprise AI platform for your own data: cited answers, tool-using agents and visual workflow automation — connect 24+ sources, keep everything behind your firewall, and trace every answer from the exact passage to the permission that allowed it.
First-party plugins — full products on the ERAG platform.
Stratum turns a legacy codebase into a validated cloud-native application through one product that deepens in four phases — with file-and-line evidence on every claim, so the plan you approve is the plan that ships.
Map every module, dependency, table and business process into one knowledge graph, scored and cited.
Retain, replatform, refactor or rearchitect — a strangler-fig sequence with a cost model and scenarios.
Characterization tests — unit, integration and golden — compiled and run in a sandbox, with a safety score.
Verified change sets, cloud artifacts (Dockerfile, Helm, Terraform) and pull requests with migration notes.
Four moves, each one useful on its own. You climb only as far as the value justifies, and you own everything at every rung.
Two to three weeks mapping your data, systems, use cases and risk — ending in a costed roadmap, not a deck of maybes.
One use case, built on real documents and real permissions, measured against a golden set before anyone calls it a success.
Into your cloud, your VPC or your air-gapped network, wired to SSO and your audit trail, with runbooks and a handover.
Evals on a schedule, cost and quality on one dashboard, and a managed team on call for as long as you want one.
Every answer cites its passage; every modernization claim cites its file and line. If a system can't show its work, we don't ship it.
Hosted, in your VPC under your own keys, or fully air-gapped with local models. Same product, same result.
Frontier models, open weights on your hardware, or both with fallbacks and budgets. No lock-in to one vendor's roadmap.
Open-source core, self-hostable, documented and handed over. There is no cliff between the engagement and the years after it.
The places where an answer without a citation is worthless, and where the system of record was written before the people using it were hired.
Wherever you allow it to. We can run the whole stack inside your VPC under your own keys, or fully air-gapped with local models on vLLM or Ollama — nothing leaves the room. On the hosted service, storage is isolated per organization and encrypted at rest. Your data and your source code are never used to train models.
No. We build on ERAG and Stratum because it is faster and because you can self-host them afterwards, but a large share of our work is on other people's stacks — Azure OpenAI, Bedrock, Vertex, LangGraph, a warehouse you already run. If the right answer is that you don't need a new platform, we'll tell you.
You keep the system, the workspace, the code and the documentation. The core of ERAG is Apache-2.0, so there is no licence cliff, and Stratum reports stay in a workspace you control. We hand over runbooks and train your team as part of every build.
Retrieval quality is measured, not assumed: golden question sets, LLM-judge scoring, a faithfulness self-check and honest abstention when the corpus doesn't contain the answer. Every claim carries a citation, and a per-query trace shows retrieval scores, fusion, reranking and the final prompt — so when something is wrong you can see exactly why.
Yes — it's usually the fastest part. SSO (OIDC and SAML 2.0), SCIM provisioning, MFA, encryption at rest, per-document access control enforced inside retrieval, moderation, audit export to your SIEM and data-residency controls are built in, and we map controls to the EU AI Act, NIST AI RMF and ISO/IEC 42001.
The audit lands in two to three weeks. A production RAG deployment on your own documents runs four to six weeks end to end. Agent automation and systems integration take six to ten. Modernization depends on the codebase, but the first archaeology report comes back in days.
Your choice. Bring your own provider keys and pay them directly, or use our hosted models — ERAG plans include AI tokens with the storage you buy. Either way you get per-role routing, fallbacks and a daily budget, so cost is a setting rather than a surprise.
Yes. The Diligence Brief is a read-only archaeology of a target's codebase delivered in days: a modernization score, the risks, key-person exposure and a modernization estimate the deal team can put in the model — with citations behind every number.
We'll ask a few questions about your documents, systems and goals, then point you to the smallest engagement that gets you there — or tell you honestly if it isn't a fit.
[email protected] · delivered remotely worldwide