# Eden AI

> Eden AI is an AI gateway. One API key, one request shape and one bill give access to the model catalogues of more than fifty providers, for both chat models and expert models such as OCR, speech, translation and image analysis.

Eden AI is a French company and serves a dedicated European endpoint that refuses, before any provider is contacted, anything not cleared for processing in the EU. This file indexes the pages worth reading. The product documentation has its own index, linked first below.

Index generated 2026-10-06. Figures in the final section carry the date they were last measured.

## Key resources

- [Documentation index](https://www.edenai.co/docs/llms.txt): llms.txt index of the full product documentation, regenerated on every docs release
- [Full documentation corpus](https://www.edenai.co/docs/llms-full.txt): every documentation page in one file
- [Documentation home](https://www.edenai.co/docs/): quickstarts, API reference and guides
- [Pricing](https://www.edenai.co/pricing): provider rates with no markup, plus a 5.5% platform fee, no subscription
- [Model catalogue](https://www.edenai.co/models): the models reachable through Eden AI, with pricing and capabilities
- [AI providers](https://www.edenai.co/providers): one page per provider: what it does well, what it costs, when to choose it
- [Integrations](https://www.edenai.co/integrations): Eden AI inside automation platforms, IDEs, agent frameworks and no-code tools
- [Blog](https://www.edenai.co/blog): benchmarks, provider launches and engineering notes
- [Dashboard](https://app.edenai.run/): create a key, watch spend, configure guardrails
- [Contact](https://www.edenai.co/contact): talk to the team

## Start here

The shortest path from nothing to a working call.

- [First LLM call](https://www.edenai.co/docs/v3/quickstart/first-llm-call.md): the OpenAI-compatible chat endpoint, in one request
- [First expert model call](https://www.edenai.co/docs/v3/quickstart/first-expert-model-call.md): the Universal AI endpoint for OCR, speech, translation and image features
- [What is an AI gateway](https://www.edenai.co/docs/v3/overview/ai-gateway.md): what a gateway does, and when you do not need one
- [LLMs vs expert models](https://www.edenai.co/docs/v3/overview/llms-vs-expert-models.md): two endpoints, and which one a given task belongs to
- [Batch OCR table extraction](https://www.edenai.co/docs/v3/quickstart/batch-ocr-tables.md): a worked end-to-end job: many documents in, one CSV out

## Routing and reliability

Choosing which provider serves a request, and what happens when one fails.

- [Provider routing](https://www.edenai.co/docs/v3/llms/provider-routing.md): name a model without a provider and let Eden AI pick by price, speed or latency, with health-aware failover
- [List LLM models](https://www.edenai.co/docs/v3/llms/listing-models.md): read the live catalogue, including capabilities and regions, at run time rather than hard-coding a model
- [Chat completions](https://www.edenai.co/docs/v3/llms/chat-completions.md): the OpenAI-compatible shape, identical across every provider
- [Responses](https://www.edenai.co/docs/v3/llms/responses.md): the stateful alternative to chat completions
- [Rate limits](https://www.edenai.co/docs/v3/overview/rate-limits.md): the per-account limit, how it is shared across keys, and how it is raised

## Cost and governance

Keeping spend and model access under control across a team.

- [Guardrails](https://www.edenai.co/docs/v3/organization/guardrails.md): deny-by-default model policies, attachable to a member, a role or a single API key, with budgets and rate limits
- [Custom API keys](https://www.edenai.co/docs/v3/general/custom-api-keys.md): per-key budgets, expiry and attribution
- [Buying credits and auto-refill](https://www.edenai.co/docs/v3/general/buying-credits.md): prepaid credits, automatic top-up, and monthly invoicing for larger teams
- [Plans and pricing](https://www.edenai.co/docs/v3/overview/plans-prices.md): no markup on provider pricing, one platform fee
- [Prompt caching](https://www.edenai.co/docs/v3/llms/prompt-caching.md): reuse stable prompt prefixes to cut latency and input cost

## Data governance and European processing

Where a request is processed, and what survives it.

- [EU endpoint](https://www.edenai.co/docs/v3/data-governance/eu-endpoint.md): same key, different base URL; anything not cleared for European processing is refused with HTTP 451 before the provider is contacted, with no silent fallback
- [Provider data policies](https://www.edenai.co/docs/v3/data-governance/provider-data-policies.md): zero data retention and no-training conditions, provider by provider
- [Data retention](https://www.edenai.co/docs/v3/data-governance/data-retention.md): what Eden AI keeps, and for how long
- [Servers location](https://www.edenai.co/docs/v3/data-governance/servers-location.md): where Eden AI and each provider host
- [Static IP](https://www.edenai.co/docs/v3/data-governance/static-ip.md): a fixed egress address for customers who allow-list

## Beyond chat

The expert model side: documents, speech, images, translation.

- [Embeddings](https://www.edenai.co/docs/v3/llms/embeddings.md): one API for embeddings, for retrieval, search and clustering
- [Structured output](https://www.edenai.co/docs/v3/llms/structured-output.md): constrain a response to valid JSON against a schema
- [Web search](https://www.edenai.co/docs/v3/llms/web-search.md): give a model access to current information
- [File upload](https://www.edenai.co/docs/v3/llms/file-upload.md): persistent file storage for documents sent to models
- [Decisions](https://www.edenai.co/docs/v3/llms/decisions.md): typed questions answered with calibrated probabilities instead of prose

## Locales

Eden AI publishes in English and French. Each locale has its own pages.

- [Eden AI in French](https://www.edenai.co/fr): the same site in French
- [Sitemap](https://www.edenai.co/sitemap.xml): every page, both locales

## About Eden AI

- Company: Eden AI, a French company, headquartered in Villeurbanne, France
- Product: an AI gateway in front of chat models and expert models, behind one API key
- Endpoints: `https://api.edenai.run` globally, `https://api.eu.edenai.run` for European processing
- Chat catalogue: 1153 model routes on the global endpoint, 286 on the European endpoint, across 35 providers (measured 2026-10-06)
- Providers whose entire catalogue is served in the EU: 7 of 35, covering 127 routes, namely greenference, infomaniak, ionos, mistral, ovhcloud, scaleway and tensorx (measured 2026-10-06)
- Expert models: 88 subfeatures across 9 families, audio, image, llm, multimodal, ocr, text, translation, video and web, served by 83 providers (measured 2026-10-06)
- Pricing: provider rates with no markup, plus a 5.5% platform fee applied when credits are purchased; no subscription and no minimum
- Regional enforcement: on the European endpoint, a request naming any model or fallback not cleared for European processing is rejected with HTTP 451 before the provider is contacted, before input validation, and without spending credit
- Access control: guardrails apply deny-by-default model policies to a member, a role or an individual API key, and are enforced server side on every request
- Default rate limit: 10 requests per second per account, shared across that account's keys, raised on request at no extra cost
- Compatibility: the chat endpoint follows the OpenAI Chat Completions and Responses shapes, so an existing OpenAI client works by changing the base URL and the key