DocsAlot turns scattered help center articles, knowledge base, and developer docs into one source of truth for humans and AI agents. It includes hosted MCP, llms.txt, and skill.md. Your docs show up in AI answers, onboarding gets faster, and agents stop reading stale context.
Maintain your help-center by asking Claude, Cursor or Codex
Create and update your knowledge base from Claude, Codex, or any compatible MCP client. Ask DocsAlot to add a page, fix an error, or improve your help center. It finds the relevant content, makes changes, creates a new version, and publishes, all from your existing workflow. See what questions users ask in your help center, discover what’s missing, and use those insights to improve your docs and decide what to build next.
separating editing from publishing is the right call. a working draft that a human explicitly promotes to the live site, with version history to roll back to, is a much saner model than letting an agent write straight to production docs. preserving frontmatter, tables, and links while switching between visual and markdown mode without breaking structure is also a harder problem than it sounds.
What needs improvement
the correction feedback loop isn't obvious from the pitch - if an agent's proposed doc update gets heavily edited or rejected by a human before publishing, does that correction feed back to the agent, or does it just try the same thing again next time?
Mintlify is more of a full hosted docs platform tied to its own build pipeline. DocsAlot's pitch is narrower and more useful for us right now: a visual editor that sits on top of docs an agent can also read and write to through MCP, without committing to a whole new hosting stack.
hey @Gal, appreciate the rating and the comments. Yes, the feedback loop back to human agent is something thats not possible at the moment, but could be a nice addition in the form of rejection event.
Would you be open for a call, would love to understand how you see this, happy to give you a few months free.
@Gal Hey Gal, with our current launch of the MCP. We have addressed your original limitation. You can now ask the agent why a particular version change was suggested. It retains history of all interactions.
Report
yo congrats faizan! the reader questions thing is lowkey the best part, do those come from the normal search too or just the ai one?
Looking real good! A lot of good things are coming out in this area, will certainly be trying it out! Does the mcp support creation for web documentations (the ones we would have in a product's site, for example?)
No, unfortunately the MCP won't have access to your source-code so it can't modify things. It only has access to the help-center hosted on DocsAlot. It can listen to source-code as a read-only thing but thats about it.
Would love to chat about what you have in-mind. Could be a good feature to add.
hey @Gal, appreciate the rating and the comments. Yes, the feedback loop back to human agent is something thats not possible at the moment, but could be a nice addition in the form of rejection event.
Would you be open for a call, would love to understand how you see this, happy to give you a few months free.
https://cal.com/faizan-khan-18shuq/30min