Integrations
GDPR-compliant models for the agents you already build
Frontière AI is an OpenAI-compatible endpoint, so any agent platform that lets you set a base URL can run on European infrastructure: change three values — base URL, API key, model ID — and on the models labeled EU sovereign the inference moves to OVHcloud or Scaleway instead of a US provider; fast-access models are served by US companies and labeled as such. Compatibility is the easy half. The half that decides whether an agent works is function calling, so we measure it rather than assume it: one real call per model, and 26 of the 26 live models measured so far returned valid tool calls.
The three values
Every integration on this page comes down to the same three settings. Whatever your platform calls them, this is what goes in:
Base URL https://getfrontiereai.eu/api/v1
API key sk-front-… (dashboard → API keys)
Model ID glm-5.2 (GET https://getfrontiereai.eu/api/v1/models)The model ID is the id field from our catalog endpoint, which is public and needs no key — point your browser at it and copy the one you want. It lists live models only, so anything you find there is callable right now.
Platform guides
Each guide gives the exact field names in the tool, the defaults that will bite you and which models to pick. A green badge means we ran it against the gateway ourselves; the others follow the vendor's official documentation, dated on the page:
n8n
Workflow automationPoint the OpenAI credential at Frontière AI and the AI Agent node runs on our catalog — on EU-hosted infrastructure when you pick a model labeled EU sovereign, tool calls included.
LiteLLM
LLM gatewayAdd Frontière AI to LiteLLM's SDK or proxy with the openai/ prefix: EU-hosted open models behind your existing virtual keys and budgets, with spend tracking and the provider of each call kept in the headers.
Microsoft Foundry
Agent platformA bring-your-own-model connection makes every catalog model selectable in Foundry Agent Service, with inference staying in Europe on the models labeled EU sovereign.
Claude Code
Coding agentPoint Claude Code at Frontière AI's Anthropic-compatible endpoint and run it on EU-hosted open models, tool calls included.
OpenCode
Coding agentDeclare Frontière AI as a custom OpenAI-compatible provider in opencode.json and code with EU-hosted open models from the terminal.
Cline
Coding agentUse Cline's OpenAI Compatible provider with Frontière AI's base URL to run the VS Code agent and its CLI on EU-hosted open models.
Kilo Code
Coding agentAdd Frontière AI as a custom provider in kilo.jsonc and run Kilo Code in VS Code, JetBrains or the CLI on EU-hosted open models.
Continue
Coding assistantAdd Frontière AI models to Continue's config.yaml with provider openai and apiBase, and use them for chat, edits and agent mode.
Aider
Coding agentRun Aider, the terminal pair programmer, on EU-hosted open models: two environment variables and the openai/ prefix on the model name.
Zed
Code editorAdd Frontière AI as an OpenAI-compatible provider in Zed's settings.json and drive the agent panel with EU-hosted open models.
goose
Desktop and CLI agentRegister Frontière AI as a custom provider in goose, desktop or CLI, and run its extensions on EU-hosted open models.
Hermes Agent
Personal agentRun Nous Research's Hermes Agent on a custom endpoint: Frontière AI's base URL in config.yaml, EU-hosted open models behind it.
OpenClaw
Personal agentAdd Frontière AI as a custom model provider in openclaw.json and run your self-hosted OpenClaw agents on EU-hosted open models.
LangChain
Agent framework (Python)Use ChatOpenAI and OpenAIEmbeddings with Frontière AI's base URL to build LangChain apps and agents on EU-hosted open models.
LlamaIndex
RAG framework (Python)Build LlamaIndex RAG pipelines and agents on EU-hosted open models with OpenAILike and OpenAILikeEmbedding pointed at Frontière AI.
Vercel AI SDK
Agent framework (TypeScript)Create an OpenAI-compatible provider for Frontière AI in the Vercel AI SDK and stream text, tools and embeddings from EU-hosted open models.
What an agent needs from a model provider
A chat integration only needs text in and text out. An agent needs more, and the difference is where most "OpenAI-compatible" endpoints quietly fall short.
First, function calling. An agent that cannot emit a tool call is a chatbot: no search, no database lookup, no action. Frontière AI relays tools and tool_choice untouched, so the answer depends entirely on the model and the provider serving it — which is why we measured it rather than assumed it. Second, streaming that still reports usage: our streams carry the usage object in the final chunk before [DONE], so a streamed agent step is billed as exactly as a blocking one. Third, patience — reasoning models think for minutes, and most agent frameworks ship with timeouts far shorter than that.
The table below is the result of the measurement, not a copy of the model cards. Each entry got one real call carrying a get_weather tool — the first batch on 6 August 2026, later models as they were measured; five models of that first batch also got the full round trip, with the tool result sent back and the final answer checked for the value it contained.
Tool calling, measured
One real call per model through the gateway, the first batch on 6 August 2026. A live model we have not measured yet is left out of this table rather than assumed to work. A refusal can come from the provider, not the model: OVHcloud answered HTTP 400, "feature 'tool calls' is not currently supported", for Qwen2.5-VL 72B (paused since 20 October 2026). The same result is machine-readable: every measured entry in GET /api/v1/models carries a tool_calling boolean (absent means not measured yet), so an agent can filter on it at startup instead of finding out in production.
Embeddings and retrieval
An agent that answers from your own documents needs two different models: one to turn text into vectors, one to write the answer. POST /api/v1/embeddings is OpenAI-compatible and serves the EU-hosted embedding models below, so with a chat model labeled EU sovereign both halves stay on European infrastructure — until now the retrieval half had to be bought elsewhere.
| Model | Dimensions | Context | Price / 100k tokens |
|---|---|---|---|
bge-multilingual-gemma2 | 3584 | 8,192 | €0.0168 |
qwen3-embedding-8b | 4096 | 40,960 | €0.0168 |
Chat and embedding models are separate catalogs: a chat model sent to /embeddings answers 404, and vice versa. GET /api/v1/models?kind=embedding returns only the valid ones. Check dimensions before you index anything — switching to a model with a different vector length means rebuilding the whole store.
What we don't serve
Two gaps worth knowing before you wire anything, because they fail as a 404 rather than as a helpful message:
| Gap | What breaks |
|---|---|
| No OpenAI Responses API (/v1/responses) | Tools that only speak the Responses API cannot connect. Frameworks that switch to it on some settings must be pinned to chat completions — LangChain: use_responses_api=False. |
| No legacy completions endpoint (/v1/completions) | Fill-in-the-middle autocomplete (Continue autocomplete, Zed edit predictions, Kilo Code autocomplete) cannot use us, nor can clients that default to it, such as LlamaIndex's OpenAILike without is_chat_model=True. |
Everything else that speaks OpenAI
No dedicated guide, but we opened each vendor's own documentation and noted where the setting lives, with the date we read it:
| Tool | Where the base URL goes | Source |
|---|---|---|
 CursorCode editor | Settings → Models → OpenAI: API Key, then the Base URL override switch. Cursor routes every request through its own servers, so prompts transit Cursor (a US company) before reaching us, and Tab completion keeps Cursor's models. | 2026-10-12 |
 Open WebUIChat interface | Admin Settings → Connections → OpenAI → Add Connection, fields URL and API Key (or OPENAI_API_BASE_URL). | 2026-08-06 |
 LibreChatChat interface | An endpoints.custom entry in librechat.yaml with baseURL, apiKey and models. | 2026-08-06 |
 Fleece AIAgent CLI | fleece setup (npm @fleeceai/cli) → a provider you name yourself (e.g. frontiere), base URL https://getfrontiereai.eu/api/v1, API shape openai, then your sk-front key. Models are addressed frontiere/<model>, e.g. frontiere/glm-5.2. | 2026-09-22 |
These lines say we located the setting in the vendor's documentation — not that we ran an agent end to end on that platform.
What this changes, and what it doesn't
Changing the base URL moves the inference. It does not move the orchestration. Your platform still runs wherever it runs — its servers see the prompts, the tool calls and the conversation history before we do, and whatever jurisdiction it sits in, it keeps. If you self-host your agent platform in Europe and call a model flagged sovereign, orchestration and inference both run in Europe (our gateway itself runs in Paris on Vercel, a US company — our privacy page says so). If your platform is a managed US service, the model is the only part this fixes.
That is a real improvement and a partial one, and it is worth telling your DPO which of the two you bought. Per-model honesty is why every entry carries a sovereign flag: models served through US infrastructure are labeled fast access, never sovereign, even when the region is European — a US company remains reachable by a US court wherever the servers sit.
FAQ
Which AI agent platforms can use GDPR-compliant models through Frontière AI?
Any platform that lets you set a custom OpenAI base URL — n8n, Microsoft Foundry, LangChain, the Vercel AI SDK, OpenCode, Cline, Hermes Agent, OpenClaw, Open WebUI, LibreChat and the many frameworks built on the OpenAI client — plus Anthropic-format clients such as Claude Code, through the /v1/messages endpoint. The models flagged sovereign run on operators with no non-EU jurisdictional control, so the inference side of your agent stays under EU law.
Do open-source models support function calling?
In our catalog, 26 of the 26 live models we have measured do — each verified by a real call rather than taken from a model card. When it fails, the cause can be the provider rather than the model: OVHcloud rejected tool calls to Qwen2.5-VL 72B outright with an HTTP 400 (the model has been paused since 20 October 2026). Function calling is what separates a usable agent from a chatbot, so check it per model rather than per provider.
Does switching the base URL make my agent GDPR-compliant?
With a model labeled EU sovereign, it makes the inference European, which is usually the hardest part to fix; a fast-access model is served by a US company, and is labeled as such. It does not change where your agent platform itself runs: that layer still processes prompts and keeps its own jurisdiction. Compliance is a property of the whole chain, so audit the orchestrator too.
Can I use Frontière AI for embeddings and RAG?
Yes. POST /api/v1/embeddings is OpenAI-compatible and serves EU-hosted embedding models, so both halves of a retrieval workflow — vectorizing documents and answering questions about them — run on European infrastructure when the answering model is one labeled EU sovereign. Chat and embedding models are separate catalogs: GET /api/v1/models?kind=embedding lists the ones valid on that endpoint.
Is there a free tier for testing an integration?
No. Frontière AI runs on prepaid credit, from €10, billed per token with no subscription — there is no free allowance. A top-up covers a lot of agent runs: the calls in the measurement on this page cost a fraction of a cent each.
Point your agents at Europe
Create an account, top up from €10, and the catalog is callable from whatever you already build in.
Create an account