The best EU-hosted LLM gateways in 2026
Seven options that keep LLM traffic inside the European Union — four managed services, one aggregation platform, and two gateways you run yourself. Compared on residency, OpenAI compatibility and published price, with every figure checked against the vendor's own page.
The short answer
An EU-hosted LLM gateway is a single API endpoint, running inside the European Union, that routes requests to multiple model providers. Pick a managed one when you want the residency guarantee without operating infrastructure; pick a self-hosted one when you already run a platform team and want the gateway on hardware you control. Both are legitimate; they fail in different ways.
The options at a glance
| Option | What it is | EU residency | OpenAI-compatible | Published price |
|---|---|---|---|---|
| 2kw.ai | Managed EU platform: OpenAI-compatible gateway, schema-driven document extraction and per-field cost tracking on one account. | EU-hosted, operated by Manfred Kunze Development GmbH in Germany. Applies to the built-in models; with BYOK your requests go to the provider you chose, under that provider's terms. | Yes — point your base URL at 2kw.ai and existing OpenAI SDK code runs unchanged. | €99 / €299 / €499 per month; annual billing costs ten months. 7-day free trial. A Custom tier is priced on request. |
| Eden AI (EU endpoint) | Managed aggregation across many AI providers — LLMs plus OCR, documents, speech, vision, translation and embeddings. | Prompts, files, requests and outputs are “hosted, processed, and routed exclusively within the European Union”. Zero data retention, DPA, SOC 2 and ISO 27001. Only providers compatible with EU residency are reachable through the endpoint. | Not stated on the EU endpoint page. | 5.5% platform fee at checkout, no subscription, no API call limit. An Advanced tier is priced on request. |
| EUrouter | Managed router to 100+ chat, embedding and reranking models. Operated by EUrouter B.V., Amsterdam. | EU-West region, EU servers, stated EU data residency. | Yes — described as “a drop in replacement for the OpenAI API schema”. | Free at €0/month (10K requests, 15% markup); Plus €39/month (1M requests, 9%); Pro €99/month (10M requests, 3%). |
| HostYourAI | Managed API for 200+ open models — Llama, DeepSeek, Qwen, Mistral, Mixtral, Gemma. Dutch operator. | Inference runs on GPUs in European data centres. Prompt and output content is not stored for training or shared with third parties; only billing metadata is retained per request. | Yes — /v1/chat/completions, plus an Anthropic-style /v1/messages endpoint. | Pay-as-you-go on prepaid credits: per token on the shared router, or per hour for a dedicated GPU. Free to start, no minimum, no fixed monthly fee. |
| LiteLLM | Open-source gateway you run yourself — an “OpenAI Proxy Server (LLM Gateway) to call 100+ LLMs in a unified interface”. | Whatever your own infrastructure is. Self-hosted and air-gapped deployment, so “your data and your keys never leave it”. The residency guarantee is one you build and operate. | Yes. | Open source, free. Enterprise adds virtual keys, budgets, spend tracking, audit logs, SSO/SCIM and RBAC at custom pricing; a 30-day trial key is offered. |
| Portkey | Open-source gateway plus a managed platform — “a unified interface for interacting with over 250 AI models”. | The managed Production tier states it is “not recommended for organizations requiring custom security controls or data residency guarantees”. Residency comes via Enterprise (private cloud / VPC hosting) or by self-hosting the open-source gateway. | Yes — documented as a drop-in OpenAI SDK replacement for Python and Node. | Open source free to self-host. Developer free (10k logs/month, “not suitable for production workloads”). Production $49/month (100k logs, $9 per additional 100k). Enterprise custom. |
| Requesty (EU) | Managed gateway — “600+ models behind one OpenAI-compatible endpoint, with routing, caching and governance included”. | “All data hosted exclusively in Frankfurt, Germany (AWS eu-central-1). No data ever leaves the European Union.” Zero data retention: “nothing is stored, logged, or cached after delivery”. | Yes — change your base_url to the EU endpoint, “zero code changes”. | “Get started free”, with custom invoicing in EUR. No per-tier figures are published on the EU page. |
Every cell was verified against the linked vendor page on 31 August 2026. Those pages are the authority — pricing and features change without notice. 2kw.ai is one of the options listed and has an obvious interest in the outcome; its row is marked so you can weigh it accordingly.
Listed alphabetically. The order is not a ranking — there is no single best option here, only options that fit different constraints.
Which one fits your situation
What EU data residency actually requires
Three separate questions get collapsed into one, and the collapse is where compliance reviews fail. First: where does the gateway itself run? Second: where do the model providers it routes to run, and under whose terms? Third: what does each party retain, and for how long? A vendor can answer the first question perfectly and still forward your prompt to a provider outside the EU.
Bring-your-own-key makes this sharper rather than softer. When you supply your own provider credentials, the request goes to that provider under your contract with them, in their regions. That is true of 2kw.ai's BYOK on Scale and Enterprise exactly as it is true elsewhere. It is a useful feature and it is not a residency guarantee.
The practical checklist: ask for a Data Processing Agreement, ask for the subprocessor list, ask what is retained after a request completes, and ask which providers are reachable under the residency promise. Requesty, HostYourAI and Eden AI all publish retention statements; read them against your own retention requirements rather than assuming zero retention is universal.
Common questions
An LLM gateway is a single API endpoint that routes requests to multiple model providers. It is EU-hosted when the gateway itself runs in the European Union and the requests it handles are processed and stored there rather than being forwarded to infrastructure outside the EU. The gateway's own hosting and the hosting of the models it routes to are separate questions, and both matter.
No. An EU-hosted gateway addresses where processing happens, which is one input to a GDPR assessment. It does not by itself cover your legal basis for processing, your data-subject rights handling, your retention policy, or the terms of the downstream model providers the gateway routes to. Ask each vendor for a Data Processing Agreement and for its subprocessor list.
Not necessarily, and this is the most common misreading. Bring-your-own-key routes the request to the provider whose key you supplied, under that provider's terms and in that provider's regions. The gateway's own residency guarantee covers the gateway, not the destination. This applies to 2kw.ai's BYOK exactly as it applies to anyone else's.
It gives you the most direct control, because the gateway runs on infrastructure you chose. It also makes you responsible for deploying, patching, monitoring and scaling it, and for the residency of the model providers you then route to. LiteLLM and Portkey's open-source gateway are both credible choices if you have a platform team; they are a poor choice if you do not.
It depends on shape, not on list price. Usage-fee models (Eden AI at 5.5%, EUrouter's 3-15% markup tiers, HostYourAI's prepaid credits) cost little at low volume and scale with spend. Flat subscriptions (2kw.ai from €99/month, Portkey Production at $49/month) are predictable and cheaper once volume is steady. Self-hosted LiteLLM has no licence cost and real infrastructure and staffing cost.
Because leaving it off would be a different kind of dishonesty. The row is marked, the ordering is alphabetical rather than ranked, and every competitor figure links to that vendor's own page so you can check it. If a competitor's entry is wrong or out of date, tell us and it gets corrected.
How this comparison was made
2kw.ai is our product and appears in the table, so read this as a vendor-published comparison. Every competitor figure was read off that vendor's own page on 31 August 2026 and is quoted or closely paraphrased from it. Where a page does not state something, the cell says so rather than inferring it. No figures were estimated, averaged or carried over from third-party articles.
Vendors change pricing and capabilities without notice, and this page is a snapshot. If an entry is wrong or has gone stale, tell us and it gets corrected.
- 2kw.ai pricing — 2kw.ai/#pricing
- Eden AI EU endpoint — www.edenai.co/eu-endpoint
- Eden AI pricing — www.edenai.co/pricing
- EUrouter — eurouter.ai/
- HostYourAI GDPR-compliant LLM API — hostyourai.com/en/product/gdpr-compliant-llm-api
- LiteLLM Enterprise — www.litellm.ai/enterprise
- LiteLLM proxy docs — docs.litellm.ai/docs/simple_proxy
- Portkey pricing — portkey.ai/pricing
- Portkey introduction — portkey.ai/docs/introduction/what-is-portkey
- Requesty EU — www.requesty.ai/eu