DeepSeek & Qwen

[AysheaK]

Problem / Motivation

Need DeepSeek & Qwen Api System Agent System in this

Proposed Solution

Need DeepSeek & Qwen Api System Agent System in this

Alternatives Considered

Need DeepSeek & Qwen Api System Agent System in this

Area

AI / Agent behaviour

Additional Context

Need DeepSeek & Qwen Api System Agent System in this

[hqhq1025]

Thanks for the suggestion!

Currently, Open Cowork primarily uses the Claude Agent SDK. However, both DeepSeek and Qwen models are accessible through OpenAI-compatible API endpoints — you can configure them in Settings → API:

1. Set Provider to Custom
2. Set Protocol to OpenAI
3. Enter your DeepSeek/Qwen API base URL and key
4. Select or type the model name

For DeepSeek:

• Base URL: https://api.deepseek.com/v1
• Model: deepseek-chat or deepseek-coder

For Qwen (DashScope):

• Base URL: https://dashscope.aliyuncs.com/compatible-mode/v1
• Model: qwen-plus or qwen-max

Native multi-provider support (including dedicated DeepSeek/Qwen integration) is on our roadmap. We're tracking this as a feature enhancement.

[github-actions]

Thanks for the suggestion. I checked the current codebase: DeepSeek and Qwen can already be used through the existing custom OpenAI-compatible provider flow, but I did not find separate built-in provider types named deepseek or qwen yet.

Relevant code:

• readme.md:49 - The project README says Open Cowork supports OpenAI-compatible APIs.
• src/shared/api-provider-guidance.ts:62 - DeepSeek is explicitly recognized as a custom provider setup with OpenAI protocol, https://api.deepseek.com/v1, and example model deepseek-chat.
• src/shared/api-model-presets.ts:101 - Built-in model presets include deepseek-chat and deepseek-reasoner.
• src/shared/api-model-presets.ts:106 - Built-in model presets also include qwen-max.
• src/shared/api-model-presets.ts:1 - Current provider types are openrouter, anthropic, custom, openai, gemini, and ollama; I did not find dedicated deepseek or qwen provider types.
• src/main/config/config-store.ts:1408 - When the app is configured for OpenAI-compatible routing, it exports OPENAI_API_KEY, OPENAI_BASE_URL, and OPENAI_MODEL for runtime use.
• src/main/claude/agent-runner.ts:1273 - The main agent runner resolves custom provider routing through the configured protocol and base URL.
• src/main/claude/agent-runner.ts:1367 - For provider === 'custom', the runtime maps auth to the selected custom protocol.
• src/tests/config/auth-utils.test.ts:27 - There are tests covering OpenAI-compatible base URLs with DashScope-style paths.

So the verified status today is:

• DeepSeek/Qwen API usage: already supported through the custom OpenAI-compatible configuration path.
• Dedicated first-class DeepSeek/Qwen provider entries: not found in the current provider list.

Need more info:

• For Qwen, do you want first-class support for DashScope OpenAI-compatible API, OpenRouter, or a self-hosted endpoint?
• Are you asking for dedicated provider presets/UI entries, or for deeper agent/runtime behavior changes beyond the current custom API support?

---

Open Cowork Bot

[hqhq1025]

Moved to Discussions → https://github.com/OpenCoworkAI/open-cowork/discussions/115

[hqhq1025]

DeepSeek and Qwen are already supported in Open Cowork:

• DeepSeek: deepseek-chat, deepseek-reasoner via OpenAI-compatible API preset
• Qwen: qwen-max via OpenAI-compatible API preset
• Local/Ollama: deepseek-r1:latest, qwen3.5:0.8b

To use them: go to Settings → API → select the OpenAI-compatible or Ollama preset, enter your API key/base URL, and choose the model.

Closing as already implemented. Feel free to reopen if you encounter specific issues with these providers.