Featurebase integration guide
How to keep Featurebase up to date
Find the outdated articles in your Featurebase knowledge base after a change to your product, before Fibi repeats them to every customer who asks.
Keeping help centers accurate at
The short answer
- What Featurebase offers
- Your Featurebase Help Center has features like Documentation analytics for ranking articles by reactions, Improve answer for fixing a weak Fibi reply, and version history for comparing edits.
- Where it stops
- Each one starts from a reaction, a bad Fibi answer or a prompt a person writes, not from your code changes.
- Where Solo fits
- Solo reads your code changes, checks them against your live and draft articles, flags what a change made wrong and drafts fixes that a person approves.
On this pageShort answer
Why it matters
Why an accurate Help Center matters
You ship a change, support hears about it a week later, and the article still describes the old screen. Meanwhile readers can't self-serve, Fibi answers from stale content, and public articles give the wrong answer in search.
Readers solve problems without a conversation
A current article ends the visit. A wrong one sends the reader into a support conversation, and a negative reaction opens a follow-up box asking what could be improved.
Fibi resolves more conversations
Fibi is trained automatically on your Help Center, Updates and feedback portal, and the Help Center's AI search answers cite your articles too. A stale article gets repeated in both places.
Public articles show up correctly in search
Search engines and AI answer engines can quote public articles, so a wrong one reaches people before anyone opens a conversation. Fixing the original later doesn't correct the copies already quoted.
Built-in tools
What Featurebase already offers for knowledge base maintenance
Featurebase has real tools for this: Documentation analytics ranks articles by reader sentiment, Improve answer turns a poor Fibi reply into an article edit, and the MCP server lets an AI tool edit articles. All three start from a reader's reaction or a person's prompt, and none of them reads your code.
| Stage | Featurebase's features | Missing features |
|---|---|---|
| 01Know upcoming changes | The GitHub integration syncs a linked issue's status to its Featurebase post, so a status change reaches your team and the people who upvoted the post. | Status syncing covers issues someone linked to a post. It doesn't read your code changes or feature flags, so a change with no linked post never reaches the person who owns the articles. |
| 02Identify outdated articles | Documentation analytics ranks articles by views and reactions with a sentiment score for each, and a negative reaction opens a follow-up box asking what could be improved. Improve answer points to the Help Center passage Fibi used in a poor answer. | Reactions and Fibi conversations show where a reader or customer already hit a wrong answer. An article that is well rated but no longer matches your product isn't flagged. |
| 03Draft updated copy | Improve answer suggests a change to the Help Center article Fibi used, which you can copy or open in the editor. The MCP server lets an AI tool such as Claude create and edit articles from a prompt. | Improve answer drafts from a conversation that already went wrong, and the MCP server drafts from whatever your AI tool is given. The server doesn't read your code, so the AI tool needs its own access to it. |
| 04Collaborate on updates | Preset roles (Admins, Managers, Contributors) set who can change workspace settings, and custom roles on the Enterprise plan include a Manage Help Center permission. Edits save automatically as a draft, so a teammate can prepare a change without touching the live article. | Featurebase documents no approval step for articles, and drafts and reactions don't say which change to your product made an article wrong, so reviewers still work out what to fix. |
| 05Publish changes | Publishing makes the current draft live, the Articles list shows Draft, Unpublished changes and Published, and bulk actions publish several articles at once. Each publish saves a version you can compare and restore into the draft. | Publishing happens when someone clicks Publish, not when your feature goes live. Version history records what changed in an article, not whether it still matches your product. |
With Solo
Featurebase Help Center management with Solo on top
Featurebase can't see your codebase, so it can't know a feature in your product changed unless someone says so. Solo syncs and indexes your live and draft Help Center articles, reads your recent code changes, and flags the articles a change made wrong. It drafts the fix, and a person on your team accepts, edits or rejects each one.
What Solo adds to Featurebase's built-in tools
| Stage | Featurebase without Solo | Featurebase with Solo |
|---|---|---|
| 01Know upcoming changes | The GitHub integration syncs a linked issue's status to its Featurebase post, so a status change reaches your team and the people who upvoted the post. | Solo reads code changes in your connected repositories, summarizes them in plain language and shows the feature flag tied to a change when one is available, so your team hears about it first. |
| 02Identify outdated articles | Documentation analytics ranks articles by views and reactions with a sentiment score for each, and a negative reaction opens a follow-up box asking what could be improved. Improve answer points to the Help Center passage Fibi used in a poor answer. | Solo compares each of your code changes with your live and draft articles and flags the ones that are confidently wrong, with a reason and the sources behind it, before readers hit them. |
| 03Draft updated copy | Improve answer suggests a change to the Help Center article Fibi used, which you can copy or open in the editor. The MCP server lets an AI tool such as Claude create and edit articles from a prompt. | Solo drafts the replacement text from your code change and shows the current and proposed wording side by side for each flagged article. |
| 04Collaborate on updates | Preset roles (Admins, Managers, Contributors) set who can change workspace settings, and custom roles on the Enterprise plan include a Manage Help Center permission. Edits save automatically as a draft, so a teammate can prepare a change without touching the live article. | Each flagged article becomes a review item with your product change behind it, where a person can edit the text, accept it, add it to a draft or reject it. |
| 05Publish changes | Publishing makes the current draft live, the Articles list shows Draft, Unpublished changes and Published, and bulk actions publish several articles at once. Each publish saves a version you can compare and restore into the draft. | Accept applies the update to the connected Featurebase article, and Add to Draft keeps it in Solo until you are ready, so your team decides when the change goes live. |
Questions
Common questions about Featurebase articles and Fibi
What teams usually ask before connecting Solo to their Help Center and Fibi.
Does Solo replace anything in Featurebase?
No. Your articles and collections stay in Featurebase, and Fibi keeps answering from them. Solo suggests changes, and the ones your team accepts are applied to the connected Featurebase article.
How do I update knowledge base articles in Featurebase after my product changes?
Open the article in the Help Center, edit it and click Publish. Your edits save as a draft until then, and each publish saves a version. The MCP server can do the same edits from an AI tool, but in both cases someone first has to know which articles the change affected.
Why does Fibi give answers that no longer match your product?
Fibi is trained automatically on your Help Center, Updates and feedback portal, along with any snippets, websites and files you add. If an article still describes last quarter's setting, Fibi repeats it. Website sources aren't recrawled automatically, so those need to be re-added.
Can Fibi suggest changes to Help Center articles?
Yes. Improve answer lets a teammate describe what went wrong in a Fibi reply, and Fibi suggests a change to the article it used. It works from that conversation, so it only knows about a change to your product after a customer has run into it.
Does the Featurebase changelog keep articles current?
No. Updates, previously called Changelog, announce your product changes to your users, and publishing one doesn't edit any Help Center article. Solo checks the articles against your code change itself.
Will Solo change Featurebase articles without a review?
No. Each suggested change waits for someone on your team to accept, edit or reject it. Solo also limits its suggestions to changes it is confident about, so it flags fewer articles rather than guessing.
Related guides
Also works with
Solo works with the help center you already run, and each guide covers what is specific to that platform.
Further reading on help center drift
Self-Updating Help Center: How It Actually Works
A self-updating help center catches product changes before customers do. Here's what "automatic" really means, and what still needs a human.
4 min read
Release Notes to Help Center Articles With AI: What Works
Piping release notes into AI for instant help articles sounds efficient, but it misses the real problem. Here's what keeps a help center accurate instead.
4 min read
How Accurate Help Center Docs Reduce Support Tickets
A specific, fixable slice of ticket volume comes from outdated help center articles. Here's how to reduce support tickets with accurate docs, without hiring.
7 min read
Find the Featurebase articles Fibi is quoting wrong
A demo runs Solo against your own Featurebase Help Center and codebase, so the inaccuracies you see are real.