How we work · English
We install AI on your tools. Ours is Claude.
A short, scoped, measurable cycle — and a Claude-native workshop: Claude Code, MCP, Skills, agents. This page exists for IT leaders, partners and recruiters who want to know exactly what is under the hood.
- 01
Diagnose
30 minutes
We isolate the real bottleneck — an operational choke point, a site that does not convert, a project that drifts. No forty-page questionnaire.
One clear hypothesis and an order of magnitude for the gain.
- 02
Scope with numbers
3 to 5 days
Perimeter, target architecture, delivery plan and a firm price. You decide on figures, not on a hunch.
A scoping document and a binding quote.
- 03
Build
3 to 10 weeks
Short sprints with a demo at every increment. You see the product move and can redirect early.
Successive releases to production — never a big bang.
- 04
Hand over & measure
Ongoing
Documentation, team training, measurement of real gains. We stay available; you are not captive.
A system your team owns, and indicators that are followed.
Our workshop
Built with Claude, every day
Everything we deliver is built with Claude (Anthropic): our daily tool at the studio and at OMNES Education, where the applications are developed with Claude Code and connected to the team's tools through MCP. We do not sell a model; we install a way of working, and we hand it over.
- Claude Code
- The workshop. We design, code and deploy in it — sites, management spaces, automations — and we install it at clients who want to own their tooling.
- MCP
- The protocol that connects Claude to the tools you already run: CRM, email, Jira, Airtable, Google Drive, banking. No platform of ours between you and your data.
- Skills
- Your context, versioned: templates, rules, voice, exceptions — written once, applied to every task, improved after every use. The asset that stays with you.
- Agents
- An always-on agent for watch, list verification and Monday reports — bridged to the workshop, proposing without ever publishing or sending on your behalf.
The rules
What we do not negotiate
AI proposes, a human validates
Nothing is sent, published or posted by a machine. It is a safety rule and a selling point.
Your context is the asset
The model is a consumable, replaceable in one line. Templates, prompts, rules and documentation are versioned and belong to you.
Personal data stays personal
Anonymised in every deliverable that leaves the studio; no personal data sent to a model when the task does not require it.
We say when it is not for you
Prices are public. We do not promise a figure we have not measured.