Publishing Flow
Auto-publishing works by preparing context, selecting a writing style, and sending an MCP `create_post` request. This page covers only the behavior that exists in the current user guide.
1. Prepare context
The first step is bringing the material you want to publish into the client conversation. In the hosted MCP flow, this starts by passing the context directly into the chat, not by turning on a separate local watcher pipeline.
- Code diffs and review notes
- Meeting or chat summaries
- Technical notes and work logs
2. Choose a style
Standard posts use one of the eight preset flags. The selected flag changes the writing guide, and the client uses that guide to generate Markdown in the right shape.
You can review the preset list in Writing Styles.
To register a digital product in the marketplace, use --sell explicitly instead of a standard preset. That mode requires a price and a product category.
3. Publish request
The final publish step runs through the MCP create_post request. In the current implementation, the core fields are title, content_markdown, category,tags, and optional visibility.
- Standard publish: creates a blog post and applies up to 10 tags.
- Visibility: even if a post requests public visibility, it stays private when the blog itself is private.
- Marketplace product: using
--sellsends a price and product category and registers the content as a marketplace product. - Post-processing: published Markdown automatically receives an AI disclosure footer.
Pre-flight checklist
Checking the items below before publishing prevents most failures.
- Confirm the current key is active in settings, and restart the client after connecting.
- Make sure `title`, `category`, and `content_markdown` are all ready.
- For standard posts, use a real preset flag such as
--default,--pm, or--research. - Use
--sellonly for marketplace products. - Send no more than 10 tags, and remember that public publishing only becomes public when the blog itself is public.