One governed interface for multi-model inference.
ELGRA is developing a neutral routing, metering and settlement layer that gives European enterprises a single, policy-controlled way to consume AI models across providers. The platform is in pre-launch development; the capabilities described on this site reflect planned architecture, not a live commercial service.
A single, governed surface for model consumption.
The planned ELGRA layer separates how an organisation consumes models from how those models are sourced. Every capability below is in design or development and is labelled accordingly.
Unified API gateway
One consistent request and response contract across heterogeneous upstream providers, so applications do not carry provider-specific integration logic.
Provider-neutral routing
Request-level selection across providers, with routing rules that remain under the customer's control rather than being fixed to a single upstream.
Failover & rate-limit handling
Designed so upstream outages and throttling are absorbed at the routing layer, avoiding retry logic in every client application.
Token metering & settlement
A consistent unit of measurement for usage and cost, with a ledger model intended to reconcile consumption across providers in one place.
Usage visibility & budgets
Consolidated views of requests, tokens, models and spend, with workspace-level budgets and alert thresholds as planned controls.
Policy & data controls
Configurable rules on provider eligibility, processing region, retention and audit logging, enforced at the routing layer on every request.
Designed for resilience, auditability and clear boundaries.
A layered design keeps policy decisions separate from provider execution, so governance rules are applied before any request leaves the platform boundary.
EU governance as a design constraint, not an add-on.
ELGRA is incorporated in Hungary. The controls below are planned product capabilities; the legal entity and its registered seat are established facts.
Items marked Planned are product intentions under development. They are not certifications held, commitments delivered, or guarantees of availability.
Indicative delivery sequence.
Phases describe intended sequencing only. No dates, service levels or availability commitments are made on this page.
Foundation
Corporate and compliance groundwork, architecture definition, and core gateway design.
Private preview
Controlled technical validation of routing, policy enforcement and metering with a limited set of participants.
Controlled rollout
Expanded access with documented operational processes, support model and commercial terms.
General availability
Self-service onboarding, published documentation and published service commitments, subject to prior phases.
Registered entity information.
The following details are reproduced in the original Hungarian registry wording, with an English reading alongside. This is the factual basis of the ELGRA entity.
This website describes a platform that is in pre-launch development. ELGRA Zrt. does not currently operate a commercial token-routing or inference service and makes no representation of existing customers, traffic volumes, model integrations, service levels or revenue.
Capability descriptions, architecture diagrams and roadmap phases are forward-looking design intentions. Any commercial or technical qualification should be assessed against separately supplied documentation and the company's registry records.
Partner enquiries
Please use the contact details supplied together with the partner application. No general contact address is published on this site.
A neutral, governed layer for European model consumption.
ELGRA Zrt. is building toward that role. Current status: pre-launch, in development. Registered in Budapest, Hungary.