# corelayer0

> corelayer0 turns any OpenAPI specification into a hosted MCP (Model Context Protocol) server in about 30 seconds — no code to write, no infrastructure to maintain. No published OpenAPI spec? corelayer0 can also generate one from the API's documentation pages. The resulting MCP URL works instantly in Claude, ChatGPT, Mistral, or any MCP-compatible client. EU-hosted (OVHcloud, France), GDPR-native: API credentials are injected server-side and never reach the LLM or leave the EU.

corelayer0 is a gateway between your APIs and every LLM: it speaks MCP on the LLM side and plain HTTP/REST on your API side. Every endpoint becomes a properly typed MCP tool, with auth, schemas and error handling preserved. The MCP URL stays stable when the spec is updated, so connected LLM clients never break.

## How it works

1. Import your OpenAPI spec — JSON or YAML, or a public HTTPS URL. OpenAPI 3.x and Swagger 2.0 supported. The spec is parsed server-side and never enters the LLM context.
2. corelayer0 generates the MCP server — every endpoint becomes a typed MCP tool; auth is injected server-side, automatically.
3. Connect your LLM — paste the MCP URL into Claude, ChatGPT, Mistral, or any compatible client. The URL stays stable across spec updates.

## Features

- Classic Mode: one MCP tool per enabled endpoint. The LLM sees every operation directly in context. Simple and explicit — optimal below ~20 endpoints.
- Code Mode: two fixed tools — search() to explore the OpenAPI spec, execute() to call the API. Constant token cost (~1,000 tokens) regardless of API size. Recommended above 20 endpoints.
- Spec import (JSON & YAML): paste a JSON or YAML spec, or point to a public HTTPS URL. OpenAPI 3.x and Swagger 2.0 supported.
- Generate from docs: no published OpenAPI spec? Point corelayer0 at the API's documentation pages (up to 5) and an AI drafts the spec. The user reviews and edits it before the project is created — nothing is saved without explicit approval.
- API token management: static Bearer token encrypted server-side, injected automatically on every outbound call. The LLM never sees credentials.
- Endpoint filtering: enable or disable each endpoint from the dashboard — expose exactly what the LLM needs, nothing more.
- Auth options: no token (public APIs), API key / static Bearer (header injection), or OAuth2 Client Credentials (machine-to-machine).
- Native OAuth 2.1 per project: each MCP server is its own Authorization Server — PKCE S256, dynamic client registration, automatic token rotation.
- EU sovereignty by design: hosted on OVHcloud (France, EU), secrets encrypted under an EU-controlled KMS. Data and credentials never leave the European Union and are not subject to the US CLOUD Act.
- Secure by default: Argon2-hashed tokens, AES-256-GCM encryption under KMS, EU-only infrastructure.

## Roadmap

- Postman Collections: import Postman collections directly, without manual OpenAPI conversion.
- Native observability: structured logs, OpenTelemetry traces, replays — see what each LLM called, with which arguments, and what came back.
- Git sync / webhooks: connect a repo or webhook; MCP tools regenerate automatically on every push, with a stable URL.

## FAQ

- Do my API data or credentials ever reach the LLM? No. Credentials are encrypted server-side (AES-256-GCM under an EU-controlled KMS) and injected on each outbound call. The LLM sees tools and results, never your tokens or raw spec.
- Which LLM clients are supported? Any MCP-compatible client — Claude, ChatGPT, Mistral, and others. You paste one stable MCP URL; it keeps working even when you update your spec.
- I don't have an OpenAPI spec — can I still use it? Yes. Point corelayer0 at your API's documentation pages (up to 5) and an AI drafts an OpenAPI spec. You review and edit it before anything is saved.
- Is my data really outside US jurisdiction? Yes. Everything is hosted on OVHcloud (France, EU) with EU-only infrastructure — no transatlantic transfer, not subject to the US CLOUD Act.

## Main pages

- [Home](https://corelayer0.com/): product overview, live demo of OpenAPI → MCP generation, how it works, features, pricing, roadmap
- [Docs](https://corelayer0.com/docs): quick start, spec formats, generate-from-docs, MCP modes, authentication (MCP + target API), managing endpoints, connecting clients (Claude, ChatGPT, Mistral, Cursor, VS Code, Windsurf), limits
- [Create an account](https://corelayer0.com/auth?mode=register): signup
- [Sign in](https://corelayer0.com/auth?mode=login): access your dashboard and MCP projects

## Legal

- [Terms](https://corelayer0.com/en/conditions): terms of service
- [Privacy](https://corelayer0.com/en/confidentialite): privacy policy (GDPR)
- [DPA](https://corelayer0.com/en/dpa): data processing agreement

## Social

- [LinkedIn](https://www.linkedin.com/company/corelayer0com): official corelayer0 company page

## Version française (résumé)

corelayer0 transforme n'importe quelle spécification OpenAPI en serveur MCP hébergé en ~30 secondes, sans écrire de code. Pas de spec OpenAPI publiée ? corelayer0 peut aussi la générer à partir des pages de documentation de l'API. L'URL MCP générée fonctionne immédiatement dans Claude, ChatGPT, Mistral ou tout client compatible MCP. Hébergé en UE (OVHcloud, France), conforme RGPD : les identifiants d'API sont injectés côté serveur et ne quittent jamais l'Union européenne.