Mode Selection
Separate API mode, AI mode, plain conversion, and advanced layout before you wire anything into a real workflow.
Mode Selection
The easiest place to get confused with md2wechat Agent API is not the endpoint list. It is the mode choice.
People often mix these into one bucket:
- API mode
- AI mode
- plain conversion
- advanced layout
- CLI / skill / raw API
They are related, but they do different jobs.
The shortest way to think about it
- If you want stable WeChat HTML, start with API mode.
- If you want stronger structure and better reading flow, add advanced layout inside API mode.
- If the result still needs another model step, then use AI mode.
API mode vs AI mode
API mode
Use it when:
- you want a stable finished result
- you want fewer steps
- you want the agent flow to be easy to validate
This is the direct production path.
AI mode
Use it when:
- you already know there is another model step after this
- you need a structured handoff, not the final output
- you accept that this is not a one-step final page
This is a prepared handoff path.
Plain conversion vs advanced layout
Plain conversion
Good for:
- short posts
- simple notices
- fast publishing
Advanced layout
Good for:
- release posts
- tutorials
- long-form argument pieces
- service pages
The point is not to make the page look richer. The point is to make the piece easier to understand and finish.
CLI, skill, or raw API
CLI
Best for local validation. Start here when you need to check conversion, preview, or discovery output first.
Skill
Best when you already work inside Claude Code, Codex, OpenCode, Claudian, or OpenClaw and want the workflow inside that runtime.
Raw API
Best when you already know you need to wire this into your own backend or automation pipeline.
A safer selection order
Follow this order:
- decide whether you need a finished output or a handoff result
- decide whether the content actually needs advanced layout
- decide whether CLI, skill, or raw API is the right entry point
Common misreads
- treating AI mode like a one-step final renderer
- using advanced layout for very short content
- assuming themes, providers, or modules before discovery
- jumping into raw API when local validation would have answered the question first