Feature proposal: add QwenCoder CLI integration

[starAndHonor]

Proposal

I'd like to propose adding QwenCoder CLI as a supported integration in @agents-dev/cli.

agents is most valuable when one .agents/ source of truth can sync MCP servers, skills, and instructions across all the coding tools a team actually uses. QwenCoder CLI seems like a strong candidate for that model, especially for teams experimenting with open-model or Alibaba/Qwen-based coding workflows.

Why this would help

Right now, teams using QwenCoder CLI alongside Codex / Claude / Gemini / Cursor / Windsurf / OpenCode still need to maintain QwenCoder configuration separately.

That reintroduces the same problems agents is solving:

• duplicated MCP setup
• drift between tools
• more onboarding friction
• extra documentation and maintenance burden

Suggested scope

An MVP could include:

• adding qwen or qwencoder as a valid integration id
• supporting agents connect --llm ... and disconnect --llm ...
• syncing MCP server configuration into QwenCoder CLI's expected format
• mapping canonical AGENTS.md instructions into whatever instruction mechanism QwenCoder CLI supports
• adding skills/workflow syncing if QwenCoder CLI has an equivalent concept
• optional status / doctor probing if the CLI supports inspection commands

Open questions

The main thing to clarify is how QwenCoder CLI expects configuration to be provided:

• project-local config, global config, or both
• file-based config vs CLI-managed config
• config format for MCP server definitions
• instruction file conventions
• whether it has a skills/workflows directory
• whether it exposes a command that can be used for health/status checks

Why it seems aligned

This feels consistent with the project goal of reducing multi-LLM configuration fragmentation. Supporting QwenCoder CLI would make agents more useful for mixed-tool teams, especially those that want to standardize around .agents/agents.json without giving up newer or less-mainstream coding CLIs.

Possible acceptance criteria

A practical first version could be considered done when:

• QwenCoder CLI can be enabled as an integration target
• agents sync materializes the expected config
• MCP server targeting works
• tests cover rendering and sync behavior
• README/docs mention the integration and any caveats

If there's interest, I'm happy to help gather the expected QwenCoder CLI config format and sketch the integration shape more concretely.

[amtiYo]

Thanks, this makes sense as an idea and the motivation is clear.
To evaluate or implement it, I’d need the concrete QwenCoder CLI config model and mapping details.
If you want to move this forward, please follow up with docs/examples for config location, MCP format, instruction handling, and any skills/workflow equivalents.