Your tools, Evanesca's models: three values to change.

The three values

Whatever your tool calls them, this is what goes in:

Base URL

https://evanesca.app/api/v1

API key

Created in your Evanesca account, sent as Authorization: Bearer (the examples read it from EVANESCA_API_KEY).

Model ID

An id from GET /models — for example gemma-4-26b

The catalogue endpoint is public: you can pick a model ID before you even have a key (model table). An ID Evanesca does not serve answers 404 model_not_found (where calls run).

Coding agents and editors

Agents that read your repository, edit files and run commands — every step a tool call, so the model you pick matters.

Coding agentClaude CodePoint Claude Code at Evanesca's Anthropic-format endpoint and run the terminal agent on open models run in Europe, tool calls included.Coding agentOpenCodeDeclare Evanesca as an OpenAI-compatible provider in opencode.json and code from the terminal with open models run in Europe.Coding agentClineChoose Cline's OpenAI Compatible provider, enter Evanesca's base URL, and the VS Code agent and its CLI run on models in Europe.Coding agentKilo CodeRegister Evanesca as a custom provider in kilo.jsonc: Kilo Code runs in VS Code, JetBrains or its CLI on models run in Europe.Coding assistantContinueList Evanesca's models in Continue's config.yaml with provider openai and apiBase, for chat, edits and agent mode on models in Europe.Coding agentAiderRun Aider, the pair programmer of the terminal, on models run in Europe: two environment variables and the openai/ prefix.Code editorZedAdd Evanesca as an OpenAI-compatible provider in Zed's settings.json and drive its agent panel with open models run in Europe.

Personal and desktop agents

Agents you run yourself, on your desktop or your own server, with their extensions and tools.

Desktop and CLI agentgooseRegister Evanesca as a custom provider in goose, on the desktop or in the CLI, and run its extensions on models run in Europe.Personal agentHermes AgentRun Nous Research's Hermes Agent on a custom endpoint: Evanesca's base URL in config.yaml, open models run in Europe behind it.Personal agentOpenClawAdd Evanesca as a model provider in openclaw.json and run your self-hosted OpenClaw agents on open models run in Europe.

Frameworks

Libraries for building your own applications, agents and RAG pipelines, in Python or TypeScript.

Agent framework (Python)LangChainPoint ChatOpenAI and OpenAIEmbeddings at Evanesca's base URL and build LangChain apps and agents on models run in Europe.RAG framework (Python)LlamaIndexBuild LlamaIndex RAG pipelines and agents on models run in Europe with OpenAILike and OpenAILikeEmbedding pointed at Evanesca.Agent framework (TypeScript)Vercel AI SDKCreate an OpenAI-compatible provider for Evanesca in the Vercel AI SDK and stream text, tools and embeddings from models in Europe.

Platforms and gateways

Workflow builders, agent platforms and LLM gateways: where the field hides, which node to use, which default gets in the way.

Workflow automationn8nOne OpenAI credential pointed at Evanesca runs your AI Agent and embeddings nodes on models in Europe. The exact fields, the Responses API switch, the timeout to raise.Agent platformMicrosoft FoundryA “bring your own model” connection makes Evanesca's models selectable in Foundry Agent Service, with the inference run in Europe.LLM gatewayLiteLLMAdd Evanesca to LiteLLM's SDK or proxy with the openai/ prefix: models run in Europe behind your virtual keys, budgets and spend tracking.

Also works with

No guide for these, but each one documents where an OpenAI-compatible base URL goes. These rows describe the tool from its documentation, not a test with Evanesca.

Tool and source	Where the base URL goes
Cursor Code editor cursor.com · read on 12 Oct 2026	Settings → Models → OpenAI: API Key, then the switch that overrides the base URL. Cursor sends every request through its own servers, so prompts pass through Cursor, a US company, before reaching Evanesca — and Tab completion keeps Cursor's models.
Fleece AI Agent CLI npmjs.com · read on 22 Sep 2026	fleece setup (the @fleeceai/cli package on npm), then a custom provider you name — evanesca, say — with https://evanesca.app/api/v1 as base URL, openai as API shape and your Evanesca key. Models are then called evanesca/<model ID>, such as evanesca/deepseek-v4.1-flash. Fleece AI is published by FLEECE AI, which also publishes Evanesca.
Open WebUI Chat interface docs.openwebui.com · read on 6 Oct 2026	Admin Settings → Connections → OpenAI → Add Connection, with the URL and API Key fields — or the OPENAI_API_BASE_URL environment variable.
LibreChat Chat interface librechat.ai · read on 6 Oct 2026	An endpoints.custom entry in librechat.yaml with name, baseURL, apiKey and models.

What an agent needs

A chat integration needs text in and text out. An agent needs three more things, and they decide whether it works:

• Tool calling. An agent acts through tool calls; a model that cannot emit them only chats. Evanesca passes tools and tool_choice through untouched, and the model table shows, model by model, whether a real tool call came back.
• Usage in streams. A streamed answer ends with a usage chunk, so a streamed agent step is charged as precisely as a blocking one (streaming).
• Time. Reasoning models can think for minutes on a long prompt: raise your platform's timeout before it cuts them off.

An agent that answers from your documents also needs an embedding model: Evanesca's run in France (embeddings and RAG).

What does not plug in

Two of OpenAI's formats Evanesca does not serve. Both fail with a 404 rather than a helpful message, so check before you wire anything.

What moves, and what stays

Changing the base URL moves the inference — the model reading your prompt — to Europe, where every Evanesca model runs. It does not move your tool: an n8n instance, a Foundry project or a hosted chat interface sees every prompt and every tool result before Evanesca does, under whatever jurisdiction it runs in.

Self-host the tool in the EU, and both halves stay in Europe. Use a managed US service, and the model is the only part this changes — worth saying plainly to whoever answers for data protection where you work. In between sit Evanesca's own servers: their host is named in the legal notice.

Frequently asked questions