We’re building Ever Works in public. Some things may be incomplete, missing, or broken while we continue improving the platform.We’re building Ever Works in public — expect a few rough edges.
Use Case
Documentation rots the moment the thing it describes changes. Ever Works generates and organizes knowledge bases and doc sites, then keeps them in sync on a schedule — so your help center, internal wiki, or product docs stay accurate without a dedicated docs team.
A coherent, navigable doc site from your source material.
Built to be easy to search, internal or public.
Re-generated on a schedule as the source changes.
Your docs are version-controlled and portable.
Each Work has a Knowledge Base the platform reads from and writes to — so your institutional context lives in your repository and powers every generation, not in a vendor silo.
A knowledge base is not a Work kind you pick from a menu. The kinds you can create today are Website, Landing Page, Blog, Directory and Awesome Repo — and every one of them carries a Knowledge Base on its Memory tab. So an internal wiki or a public help center is a Website or Blog Work whose pages are generated from the knowledge you load into that workbench, which is also the knowledge every scheduled run reads before it writes a word.
Four screens, in order. The path is the same whether the result stays behind your login or ships as a public help center.
Press + New in the sidebar, describe the docs site or help center you want, and pick the Website or Blog chip — work-kind chips hand off to /works/new, where you name it and choose a website template. There is no knowledge-base chip to look for, because the knowledge lives inside the Work you just made.
Open the Work, switch to the Memory tab and drag files onto the tree. PDF, Word, Excel, PowerPoint, CSV, TSV and HTML are converted to Markdown in process, call and meeting recordings are transcribed into a transcripts document, and every original shows its progress as Pending → Extracting… → Extracted. Web pages arriving as source URLs are handled by the default content extractor, and Notion pages once you enable the Notion extractor at Settings → Plugins → Content Processors.
The workbench has a KB-scoped AI panel: mention a document as @kb:brand/voice to pin it into context, and answers come back with citations to the documents that produced them. Anything an Agent writes back queues at /works/:id/kb/review, where you Accept, Edit then accept, Supersede or Archive it — and a lock, full or additions-only, keeps a settled page from being rewritten by the next run.
On the Worker tab, open Schedule and pick one of the seven cadences from hourly to monthly; Run now fires one immediately. Each run regenerates the pages from the knowledge you have now, and the History sub-tab records what changed — so the published site tracks its sources instead of drifting away from them.