Skip to content

Compatibility guide

BrokenGPT connection for BrokenCode

Understand the hosted BrokenGPT connection used by BrokenCode, including authorization, stable model aliases, usage, and privacy.

UPDATED 03 Aug 20264 min readTECHNICAL GUIDE
01

One account connects the coding client to hosted inference.

The legacy /docs/zen URL now documents the BrokenGPT connection. BrokenCode authorizes through your BrokenGPT account and sends valid prompts to the configured privately routed open-model service.

02

Connect, verify, and return to the project.

  1. Choose BrokenGPT from Model connections.
  2. Complete browser authorization.
  3. Return to BrokenCode and confirm the green connected state.
  4. Start or resume the project session.
03

The public contract is explicit without exposing routing internals.

  • Requests reach configured inference; there is no fake local responder.
  • Routing uses a private mixture of open-source models.
  • Provider and checkpoint identities are not exposed through the client.
  • Disconnecting removes the local account link from BrokenCode.

STRAIGHT ANSWERS

Frequently asked questions

01Why does this URL still say zen?

It is retained as a compatibility URL for installed BrokenCode clients and now points to the current BrokenGPT connection guide.

02Which model name should integrations use?

Use the stable public alias broken-one.

Authorize

Connect the client and continue coding.

Complete device authorization with the same BrokenGPT account you use for chat and API access.

Connect BrokenCode