Need an example of basic OpenAI Compatible custom model

[ChristopherJohnson25]

I'm having a difficult time setting up a basic OpenAI compatible custom model. I have no issues on other platforms like Cline/Continue.dev. I'm likely missing some piece of config. Models fetch works, so connection is successful but when doing completion, I get 500s.

Example:

endpoints: custom: - name: "CustomTest" apiKey: "1234" baseURL: "https://endpoint.com/v1" addParams: max_tokens: 2048 headers: x-api-version: "beta" x-provider: "1234" x-region-scope: "us" x-user: "tester" x-http-token: "1234" models: default: ['anthropic.claude-sonnet-4-20250514-v1:0'] fetch: true titleConvo: false titleModel: 'current_model' directEndpoint: false summarize: false summaryModel: 'current_model' forcePrompt: false modelDisplayLabel: 'CustomTest' iconURL: https://images.crunchbase.com/image/upload/c_pad,f_auto,q_auto:eco,dpr_1/rjqy7ghvjoiu4cd1xjbf

[ChristopherJohnson25]

I figured this out, I needed dropParams: - "user"

As followup, I have a header which requires user name, is it possible to dynamically include this in the yaml?

[ChristopherJohnson25]

Additional follow up, this custom endpoint also fails when any MCP server is added. Using the same MCP server with xAI, for example, works with no problem. Does a custom endpoint require some specific MCP config? The same endpoint with the same MCP Server works on Cline.

[ChristopherJohnson25]

closing this as original issue is resolved, will open another re: MCP.