ToyGate logoToyGate

Integrations

Setup guides for the clients we've personally verified — standard API SDKs, AI CLIs, and frameworks.

Docs

Any client that speaks the standard chat-completion protocol works against the gateway. The guides below are the ones we've personally verified — copy/paste a config block, set a gw_ key, you're done.

If your tool isn't listed, the recipe is always the same:

  1. Set the standard API base URL to your gateway's /v1 endpoint.
  2. Use a gw_… key as the API key.
  3. Use a model slug from the catalog — never an upstream model id.

Verified guides

Troubleshooting

If a tool refuses your key with a generic "auth failed":

  • Did you keep the gw_ prefix? Some tools trim it expecting sk-….
  • Is the base URL exactly <GatewayUrl/>/v1 (note the /v1 suffix for OpenAI-compatible clients)?
  • Is the model slug enabled in the catalog? Disabled models return 404, which some tools surface as a generic auth error.

If responses arrive but tool-calls don't work, the upstream may not support function calling for that model. Check the model's capabilities on the catalog page — capabilities.tools = true means the gateway has verified function-call support end-to-end.