> For the complete documentation index, see [llms.txt](https://help.cubelet.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.cubelet.ai/platform-and-integration/connecting-a-client/auth-and-access.md).

# Auth & access

*How authentication works across the cubelet.ai MCP fleet.*

## Every server requires an authenticated request

Fleet servers reject unauthenticated requests. The credential is a platform-issued **bearer token**, presented in the `Authorization` header on every request:

```
Authorization: Bearer <token>
```

Tokens are issued by the platform team — request one through your cubelet.ai contact. Keep the token out of source control and client config files; reference it from an environment variable instead (the Claude and Cursor quickstarts show how).

## Access is per server

Credentials and access policies are managed per server. Don't assume a credential issued for one fleet server is valid for another — confirm you're provisioned for the specific server you need from the MCP servers page before connecting.

## Requests are logged

Tool requests to the fleet are logged server-side — which server handled the request and which tool was invoked. On a platform whose product is a verifiable credential (see the Glossary entry on Passport), that record is part of the provenance story: a durable trail of who called what, when — for every agent and integration touching the platform's tools, not just for learners earning a credential.