We deploy the engine.
You run it.
A scoped engagement that takes 8 to 12 weeks. At the end, the source code is in your remote, your team knows how to operate it, and an autonomous worker is running against your real data.
From scope to handover. Five beats.
Roughly 8 to 12 weeks end to end, with weekly demos against your real data. The retainer afterwards is optional — most clients keep it.
Scoping workshop
Two days on-site. We map your data flows, your tools, your residency constraints, your security posture. Output: an agent-graph proposal and a deployment topology.
Build & integrate
We configure the engine for your problem. Your tools wired in. Your documents indexed. Iterating against your real data, with weekly demos.
Deploy on your hardware
On-prem cluster, your VPC, or air-gapped. We bring the runtime, you keep the keys. Cutover with parallel run; rollback plan in place.
Hand over & train
Source code in your private remote. Two days of training for your team. Q&A archive. After this, you can run it without us — most clients keep us on retainer anyway.
Support retainer
Optional. Monthly retainer for upgrades, new agents, model swaps, eval expansion. SLA on Slack, scheduled office hours.
Who this is for.
Teams with sensitive data or critical processes — where sending a prompt to a third-party API isn't an option, and where the cost of a wrong answer is more than a refund.
Industrial · MES
Plant-floor copilots that read PLC traces, write maintenance tickets, and never leave the production network.
Legal · Compliance
Document workflows over sealed corpora. Automated review with full audit trail and human-in-the-loop on every decision.
Finance · Treasury
Reconciliation and analyst-grade workflows on internal data, with deterministic eval and explicit kill switches.
Healthcare · R&D
GDPR-bound, on-prem agents over patient or trial data. Every byte stays inside your network.
Public sector
Citizen-services intake and triage on sovereign infrastructure. FR-language pipelines, no cross-border egress.
Engineering teams
EDA, simulation, codegen — long-running optimisation loops on your hardware, where iteration speed is the moat.
Tell us where the data can't leave.
We'll bring the agents to it.
Thirty minutes, entirely free. We'll walk through your problems, your residency constraints, and where the autonomous worker would actually move the needle.