The problem
AI coding exposes your most valuable asset
Cloud assistants send source code, secrets and know-how to US hyperscalers. In regulated environments, that's a no-go.
European Sovereign Agentic Coding Appliance
Agentic coding without data leakage.
On-premises. Air-gapped. Sovereign in Europe.
ESACA brings agentic coding into regulated enterprise environments, so your source code, secrets and know-how stay inside your own house.
The solution: ESACA
Agentic coding, fully in-house
ESACA runs agentic coding on-premises and air-gapped in your own data centre — sovereign models, turnkey.
Your advantage
AI speed, zero loss of control
Full compliance (NIS-2, BaFin, ISO 27001, GDPR), complete data sovereignty and no vendor lock-in — provable with the Benchmark.
Let's explain ESACA
In the video, Jan Jikeli, Founder of ESACA, explains the project during his breakout session at the Fsas Technologies Summit 2026.
The video is in German.
Our mission
We help organisations build and run their digital business model
— sovereignly, and with AI.
Coding is the starting point, sovereignty a requirement, and adoption happens step by step.
What is ESACA
Agentic coding that stays inside your house
ESACA — European Sovereign Agentic Coding Appliance — brings agentic coding into regulated enterprise environments. It runs fully inside your own data centre: no data egress, no external API calls, no hidden telemetry.
Coding is the core. ESACA gives your developers a production environment for coding agents on real projects. And because it runs safely, guided and under your control, it also enables people across the organisation to work with AI sovereignly.
In short: agentic coding power — without your IP ever leaving the building.
E
European
Hosted and run in Europe.
S
Sovereign
Fully under your control.
A
Agentic
Autonomous coding agents.
C
Coding
Built for real projects.
A
Appliance
Ready to run in your environment.
The platform
Build and prove: Workbench and Benchmark
ESACA starts with two tightly integrated tools: the Workbench and the Benchmark. The sovereign model infrastructure runs underneath, and nobody has to operate it.
Build
Workbench
A production environment for coding agents on real projects, with a binding review flow and audit trail. Human-in-the-loop stays binding.
- Runs on your sovereign, on-premise models
- Bitbucket and GitLab integration
- Sandboxed execution, human-in-the-loop
Prove it holds up
Benchmark
An evaluation framework that measures how well models perform on your own internal coding challenges.
- Six scoring dimensions, reproducible runs
- Runtime, static analysis and LLM-as-a-judge metrics
- Your own scenarios, on your own hardware
The status quo
Cloud coding assistants are a risk in regulated environments
01
IP leaves the building
Source code, secrets and architectural know-how are sent to US hyperscalers, typically outside your own contractual framework.
02
Compliance conflict
NIS-2, BaFin/MaRisk, ISO 27001 and GDPR requirements cannot be reliably evidenced with SaaS coding tools.
03
Ban or shadow IT
The result: blanket bans slow development down, or teams use tools behind IT's back. Both are costly.
On-prem models are hot on the heels of the cloud
Benchmark results of on-prem and cloud models over time.
What benchmarks are tracking right now is, above all, how quickly open models catch up with the proprietary frontier. Every leap by a cloud model is matched by on-prem models within a few months — for two years the lag has mostly stayed below ten percentage points, widening only briefly after major frontier releases. And around 71 % of real-world LLM queries could already be answered locally today.*
Running on-premises is therefore no longer just a compliance obligation — it is increasingly worth it in its own right: full data sovereignty, predictable costs and no vendor lock-in, with results close to the cloud frontier. ESACA brings exactly these models to your data centre, turnkey — and with the Benchmark you can measure at any time which model is sufficient for your tasks.
Why benchmark at all?
Before a single euro goes into GPUs, the Benchmark turns guesswork into evidence: it measures candidate models on the work your teams really do, not on a public leaderboard a vendor may have trained against.
Choose what to self-host
Measure candidate models on your own coding challenges and see which one earns a place on your infrastructure.
Explain the gap between teams
See why a model shines for one team and frustrates another by testing each team's real scenarios and stacks.
Re-test on every release
Define your scenarios once and replay them automatically whenever a new model appears, so an upgrade becomes a measured decision.
Promote with confidence
When a model proves itself, switch it straight into the Workbench.
The approach
Three pillars of sovereign coding
01
Air-gapped appliance
Runs fully offline inside your own data centre. No data egress, no external API calls, no hidden telemetry channel.
02
Sovereign models
European LLMs on your own hardware. Updates via a controlled supply chain: you decide what is rolled out, and when.
03
Audit-ready by design
A full audit trail of every agent action. RBAC, SSO and SIEM integration address NIS-2, ISO 27001, GDPR and BaFin requirements.
Why it matters
On your premises, under your control
Four things ESACA keeps firmly in your hands.
01
AI on your own premises
Run AI on your own infrastructure — on-premises or fully air-gapped. The appliance operates inside your own data centre, not someone else's cloud.
02
Data stays in the company
Source code, secrets and architectural know-how never leave the building. No egress to external clouds, no hidden telemetry.
03
Independence from AI providers
European models on your own hardware mean no lock-in: no forced upgrades, no surprise price changes, no provider deciding your roadmap.
04
Safe, sovereign enablement
People across the organisation can work with AI safely and sovereignly — guided, with human-in-the-loop and a full audit trail.
Who it's for
Built for regulated, high-stakes environments
01
Critical infrastructure & utilities
Organisations that may use cloud in places, but cannot rely on it everywhere, and need an alternative they fully control.
02
Public sector & administration
Teams with recurring processes and strict data boundaries that want to automate safely without sending data abroad.
03
High-security & regulated industries
Environments — from defence to legal to healthcare — bound by NIS-2, BaFin, ISO 27001 and GDPR, where AI must stay inside their own data centre.