A-Gentic
Contact Book a demo

Hosted and operated in the European Union

Run your AI agents on European cloud

A-Gentic executes your agents in isolated sandboxes on European infrastructure, with strict separation between tenants. One API, your connectors, your models — and no data leaving the EU.

Book a demo See the platform
No card required · Deployed in an EU region · Full data portability

acme-eu · Runs

eu-west-3 · Paris

run_8f21c

support-tier-1 · mistral-large · 00:12 · €0.031

Running

run_4a02c

invoice-triage · mistral-small · 00:48 · €0.014

Completed

run_1d77e

weekly-report · routine · scheduled 08:00

Queued
Residency EU only Tenant acme-eu Sandbox ephemeral
  • EU data residency
  • Encryption in transit and at rest
  • Article 28 processing agreement
  • Per-tenant audit logs

Why A-Gentic

Agent infrastructure that passes a procurement review

Three things every European buyer asks about before a single agent reaches production.

Multi-tenant by design

Organisations, members and sub-accounts. Every agent, secret and run belongs to one tenant, enforced by row level security down to the database and by separate sandboxes at run time.

European providers

Mistral, Scaleway, OVHcloud, Infomaniak, Exoscale — or your own OpenAI-compatible gateway. Compute and models are chosen per organisation, never imposed by us.

GDPR from day one

Processing agreement under Article 28, published sub-processors, configurable retention, export and erasure by data subject, and an exportable audit trail.

How it works

From account to production in three steps

No infrastructure to provision. You pick a region, connect your tools, and call the API.

Create an organisation

Pick the European region it runs in and invite your team. Members and sub-accounts inherit the tenant boundary automatically.

Connect your tools

OAuth is handled end to end. Tokens are encrypted in a per-tenant vault and refreshed for you. Internal systems connect over MCP.

Launch runs

Call the API, schedule a routine or trigger on a webhook. Every run is queued, isolated, recorded and capped in cost.

The platform

What it takes to run agents in production

Not another framework. An execution platform: queueing, isolation, connectors, scheduling and an audit trail.

Isolated sandboxes

Each run starts in an ephemeral machine with filtered networking and a bounded lifetime. Nothing persists between two agents.

Connectors and MCP

Mail, calendar, storage, CRM and your internal services, exposed to agents through managed OAuth and MCP servers.

Scheduled routines

Trigger an agent on a schedule, a webhook or an event. Every execution is recorded, replayable and capped in cost.

Full traceability

Inputs, outputs, tool calls, cost per run and per organisation. History kept for exactly as long as your retention policy says.

Queue and quotas

Concurrency limits, retries with backoff and per-organisation quotas, so one tenant can never starve another.

REST API and console

Everything the console does is a public endpoint: organisations, agents, runs, outputs, logs and OAuth connectors.

Sovereignty

Your data stays in Europe

Compute, inference, storage and logs all live in European regions, operated by companies under European law. No silent failover to a region outside the EU.

  • Residency pinned per organisationThe region is a tenant setting, not a best-effort preference.
  • Outside US CLOUD Act reachEuropean operators, European ownership, European jurisdiction.
  • No training on your dataExcluded in our terms and required contractually from every model vendor we use.
  • Bring your own modelOpenAI-compatible gateway, self-hosted weights or a dedicated instance at your own provider.
POST /v1/runs
# Start an agent in a European sandbox
curl -X POST https://api.a-gentic.eu/v1/runs \
  -H "Authorization: Bearer $AGENTIC_KEY" \
  -H "X-Org: acme-eu" \
  -d '{
    "agent":    "support-tier-1",
    "region":   "eu-west-3",
    "provider": "mistral",
    "input":    "Work through today's queue",
    "tools":    ["mail", "calendar", "crm"]
  }'

202 Accepted
{ "run_id":    "run_8f21c",
  "status":    "queued",
  "residency": "eu" }

Providers

European vendors, selected per organisation

Change provider or region without migrating your agents, your connectors or your history.

ProviderCountryUsed forRegions
Mistral AIFRLanguage modelsParis
ScalewayFRCompute, GPU, inferenceParis, Amsterdam, Warsaw
OVHcloudFRCompute, object storageGravelines, Frankfurt
InfomaniakCHInference, storageGeneva
ExoscaleCHCompute, sandboxesZurich, Vienna
HetznerDEDedicated computeNuremberg, Helsinki

Compliance

GDPR treated as a feature, not a checkbox

A-Gentic acts as a processor under Article 28. The agreement, the technical measures and the tooling to honour data subject rights are available on day one.

Processing agreement

DPA signable online, sub-processor list published, changes notified before they take effect.

Data minimisation

Retention configurable per organisation, down to zero retention of agent inputs and outputs.

Data subject rights

Export and erasure by subject, propagated to runs, logs and attachments.

Encryption

TLS in transit, encryption at rest, secrets and OAuth tokens held in an access-audited vault.

Audit log

Who ran what, with which agent, on which data, through which connector. Exportable for your own audits.

Impact assessment

DPIA template provided, with a data flow map of the platform and its processing activities.

FAQ

Questions we get asked

Where do my agents actually run?

In the European region set for your organisation: Paris, Frankfurt, Amsterdam, Zurich or Warsaw depending on the provider. The region applies to execution, storage and logs alike.

Can I use a non-European model?

Yes, and it is an explicit per-organisation decision. As long as no non-EU provider is enabled, no data leaves Europe. The console flags every organisation that has stepped outside the EU perimeter.

How is one tenant isolated from another?

Each organisation gets its own logical space, enforced by row level security in the database and by separate ephemeral sandboxes at run time. Secrets live in a per-tenant vault and are never shared across organisations.

What happens if we leave?

You export agents, configurations, connectors, run history and associated data in open formats. On request, complete erasure is carried out and attested within 30 days.

Is our data used to train models?

No. It is excluded in our terms and contractually required of the European model vendors we work with.

Is there an API?

Yes. Everything the console does goes through a public REST API: organisations, agents, runs, outputs, logs and OAuth connectors.

Your agents, your infrastructure, your jurisdiction

A 30-minute walkthrough on your own use case: a real agent, in a European sandbox, wired to your tools.

Book a demo Request the documentation
Reply within one business day · contact@a-gentic.eu