Please add "OpenAI Compatible" and "Anthropic Compatible" provider configuration

[complyue]

Please explain the motivation behind the feature request.

Many LLM api providers provide OpenAI/Anthropic compatible endpoints, with their own BASE_URL and model names, e.g.

https://docs.z.ai/scenario-example/develop-tools/roo
https://docs.bigmodel.cn/cn/guide/develop/roo

https://docs.z.ai/scenario-example/develop-tools/claude
https://docs.bigmodel.cn/cn/guide/develop/claude

Currently goose has fixed provider cfg items, not working with those broader LLM api providers yet.

Describe the solution you'd like

Please enable the user of goose to create custom providers those either OpenAI or Anthropic compatible, with custom BASE_URL and API_KEYs, and ideally with custom model list for the user to select from.

Describe alternatives you've considered

N/A

Additional context

Cline, Roo Code and Kilo Code all support "OpenAI Compatible" providers for long time, and worked really while.

• I have verified this does not duplicate an existing feature request

[shali1995]

same cant use this at all, please add this already. this is absolutly anoyying. do it like roo or cline..... oh my god............... cna you guys please get connected to the people andundertadn how we use it really?

[taniacryptid]

Thank you for posting this. Would love to get this working in theory, after immediate bug fixes and resolves. The main thing I know off-hand is that the team will want to ensure custom endpoints added here work when implemented. I'm sorry you can't use it right now via OpenRouter (get that it's your personal preference) and leaving this comment to bump the thread!

[DavidSint]

Just bumping this as it'd be great if one could use goose with the Vercel v0 family of models which is compatible with the OpenAI Chat Completions API spec.

[richardstevenhack]

Obviously. Every new tools seems to start by "we support these providers" - rather than "we support YOUR providers."
It's called "standardization".

[Developerayo]

Been looking into this, should push a pr soon with the adding of custom Anthropic/OpenAI providers

[complyue]

I see #4099 merged and released as with 1.6.0, but no luck to get it work:

~/.config/goose/custom_providers/custom_bigm_-_glm.json

{
  "name": "custom_bigm_-_glm",
  "engine": "anthropic",
  "display_name": "BIGM - GLM",
  "description": "Custom BIGM - GLM provider",
  "api_key_env": "CUSTOM_BIGM_-_GLM_API_KEY",
  "base_url": "https://open.bigmodel.cn/api/anthropic/v1/messages",
  "models": [
    {
      "name": "glm-4.5",
      "context_limit": 128000,
      "input_token_cost": null,
      "output_token_cost": null,
      "currency": null,
      "supports_cache_control": null
    },
    {
      "name": "glm-45.-air",
      "context_limit": 128000,
      "input_token_cost": null,
      "output_token_cost": null,
      "currency": null,
      "supports_cache_control": null
    }
  ],
  "headers": null,
  "timeout_seconds": null,
  "supports_streaming": true
}

% goose session
starting session | provider: custom_bigm_-_glm model: glm-4.5
    logging to /Users/cyue/.local/share/goose/sessions/20250826_183556.jsonl
    working directory: /ws/AiWorks/goose-poc

Goose is running! Enter your instructions, or try asking what goose can do.

Context: ○○○○○○○○○○ 0% (0/128000 tokens)
( O)> hi
  2025-08-26T10:35:57.785345Z  WARN goose::providers::utils: Provider request failed with status: 404 Not Found. Payload: Some(Object {"detail": String("Not Found")}). Returning error: RequestFailed("Request failed with status: 404 Not Found")
    at crates/goose/src/providers/utils.rs:110

  2025-08-26T10:35:57.785432Z ERROR goose::session::storage: Failed to generate session description: Request failed: Request failed with status: 404 Not Found
    at crates/goose/src/session/storage.rs:1299

---

update: added a trailing / to real base url so "base_url": "https://open.bigmodel.cn/api/anthropic/" does work now ...

[complyue]

Thanks @Developerayo for so quick response! ❤️

Yet another provider fails too, even with trailing / on base_url:

~/.config/goose/custom_providers/custom_ark_-_deepseek.json

{
  "name": "custom_ark_-_deepseek",
  "engine": "openai",
  "display_name": "Ark - DeepSeek",
  "description": "Custom Ark - DeepSeek provider",
  "api_key_env": "CUSTOM_ARK_-_DEEPSEEK_API_KEY",
  "base_url": "https://ark.cn-beijing.volces.com/api/v3/",
  "models": [
    {
      "name": "deepseek-v3-1-250821",
      "context_limit": 128000,
      "input_token_cost": null,
      "output_token_cost": null,
      "currency": null,
      "supports_cache_control": null
    }
  ],
  "headers": null,
  "timeout_seconds": null,
  "supports_streaming": true
}

 % goose session
starting session | provider: custom_ark_-_deepseek model: deepseek-v3-1-250821
    logging to /Users/cyue/.local/share/goose/sessions/20250826_202608.jsonl
    working directory: /ws/Games/VecRacer

Goose is running! Enter your instructions, or try asking what goose can do.

Context: ○○○○○○○○○○ 0% (0/128000 tokens)
( O)> hi
  2025-08-26T12:26:14.424467Z ERROR goose::session::storage: Failed to generate session description: Request failed: error decoding response body
    at crates/goose/src/session/storage.rs:1299

Can I have goose do some verbose logging to help diagnosing it?

[Developerayo]

Happy to help @complyue 🫡

The error you got probably happened cos OpenAI expects the full path, OPENAI_BASE_PATH uses exactly what was provided, in your case, you have https://ark.cn-beijing.volces.com/api/v3/ but it expects the completions path as well /chat/completions === https://ark.cn-beijing.volces.com/api/v3/chat/completions

For verbose logging lots of options, all here:
https://github.com/block/goose/blob/main/crates/goose-cli/src/logging.rs#L117-L130

For your case tho you could use RUST_LOG=goose::providers=trace goose session, it'll return the url, payload and resp

[complyue]

Yeah! "base_url": "https://ark.cn-beijing.volces.com/api/v3/chat/completions", does the trick! It works now 👍

Tho verbose logging does not fires:

cyue@cymp goose-poc % RUST_LOG=goose::providers=trace goose session
starting session | provider: custom_ark_-_deepseek model: deepseek-v3-1-250821
    logging to /Users/cyue/.local/share/goose/sessions/20250826_215645.jsonl
    working directory: /ws/AiWorks/goose-poc

Goose is running! Enter your instructions, or try asking what goose can do.

Context: ○○○○○○○○○○ 0% (0/128000 tokens)
( O)> hi
  2025-08-26T13:56:47.099491Z ERROR goose::session::storage: Failed to generate session description: Request failed: error decoding response body
    at crates/goose/src/session/storage.rs:1299

cyue@cymp goose-poc % 
cyue@cymp goose-poc % 
cyue@cymp goose-poc % RUST_LOG=goose=trace goose session           
starting session | provider: custom_ark_-_deepseek model: deepseek-v3-1-250821
    logging to /Users/cyue/.local/share/goose/sessions/20250826_215655.jsonl
    working directory: /ws/AiWorks/goose-poc

Goose is running! Enter your instructions, or try asking what goose can do.

Context: ○○○○○○○○○○ 0% (0/128000 tokens)
( O)> hi
  2025-08-26T13:56:57.108797Z ERROR goose::session::storage: Failed to generate session description: Request failed: error decoding response body
    at crates/goose/src/session/storage.rs:1299