Skip to content
HelpWell

Turn an unanswered question into a help article

List the gaps, write an article for the most-asked one, then mark the gap as resolved. A gap is a question the support agent could not answer from your documentation. The list is your backlog of articles to write, most-asked first.

The steps below use the HelpWell MCP server, which your AI coding tool calls.

  1. #List the gaps

    Call list_gaps. By default it returns gaps with the status "new", the ones nobody has handled yet. Pass status to see gaps in another status.

  2. #Check that the answer doesn't already exist

    Call search_articles with the question in the customer's words. Use read_article with the slug to read a hit in full.

  3. #Pick a collection

    Call list_collections to see each collection's id, title, description and article count. The id is the folder name under the docs directory and the first segment of every slug. Use it to decide where the new article belongs.

  4. #Write the article

    Call create_article with a collection, a title and a body. The body is Markdown and must not repeat the title as a heading. You can also pass a description (one sentence shown under the title) and a slug. The slug defaults to a name derived from the title.

  5. #Mark the gap as handled

    Call resolve_gap with the gap id from list_gaps and the articleSlug of the new article.

#What to know about create_article

  • The article is saved as a draft unless you set status to "published".
  • A published article needs a body. A draft may have an empty one.
  • It fails if the file already exists.
  • The article is indexed straight away, so search_articles and resolve_gap see it.
  • If OPENROUTER_API_KEY is set in the server environment, the article is also embedded for semantic search. The result says "embedded".
  • A person reviews the file like any other change.

To change an article later, use update_article with its slug. Only the fields you pass change. An article that would end up published with an empty body is refused.

Important

Call resolve_gap only after the article exists and has been indexed.

Was this article helpful?