02 · Knowledge
One knowledge graph. Three audiences. No copy-paste.
The same source of truth serves your public help center, your signed-in customers, and your internal runbooks — with enforcement at retrieval, not at render.
How it works
From a capture to a governed article, without a second write-up.
- 01A capture becomes an articleOne click promotes a guide into a knowledge base article, keeping its steps, its screenshots and its link back to the capture.
- 02Tag it once per audiencePublic, customer, internal. A chunk can carry several; every query carries a scope. You write once and it appears where it should.
- 03Publish to your own domainA branded help center, on your own domain with custom CSS from Scale. Guides keep a full version history you can diff and restore.
- 04Freshness gets an ownerContent can carry a review interval and a named owner. Past its date it is flagged, and a content.verification_due event can drive a workflow or webhook. The screen for setting owners is coming soon.
In the product
Your help center, on your domain
The public help center on your domain, with customer and internal content kept out of it at retrieval. The public reader never learns the internal guide exists.
help-center.pngBranded help center on a custom domain
The detail
Governance that holds up when someone forwards a link they should not have.
Enforcement at retrievalEvery chunk is audience-tagged and every query carries a scope, so an internal runbook cannot surface in a public answer even if someone links to it directly.
Leak probes in CICross-audience retrieval probes run on every build, and a nightly job re-checks what is cleared for public. The internal answer appearing publicly is a build failure, not an incident report.
Guide exportsMarkdown, HTML and Word, landing in a downloads queue with an email when ready — useful when a customer’s procurement wants the runbook as a document. PDF is built and waiting on its renderer; SCORM is not built.
Versioning with rollbackEvery guide change is a version you can diff and restore from the editor. Article versions are recorded too; their history screen is coming soon.
Machine translation — coming soonA machine-translated article goes to review marked as machine output, and is never published until a person signs off. The engine is built; the editor screens for it are not yet.
Retrieval API and MCP serverYour other AI tools can query the same audience-scoped knowledge, so the knowledge base stops being a dead end. The MCP server is one you run yourself; its npm release is coming soon.
Where this sits
Knowledge is the asset the two return paths keep improving.
01Capture
02Knowledge
03Engage
04AI agent
Capture flows to Knowledge, to Engage, to the AI agent; a miss becomes a work order and a resolution becomes a draft, feeding back into Knowledge.
Honest limits
Some of this is gated, and translation is still translation.
You will find this section on every feature page. If a constraint would annoy you in week two, we would rather you knew in minute one.
- Machine translation is machine translation, and it is not in the editor yet. When it lands, a machine draft still goes to review and is never published until someone signs off.
- Custom CSS and a custom domain start on Scale. The hosted help center itself starts on Starter.
- Public-publish approval — an explicit approval and a redaction check before anything goes public — is Enterprise only. Today the author can give that approval; a separate-reviewer rule is not built.
Next
04 · AI agentIt cites its sources. Or it declines.You have been burned by a confident chatbot. The behaviour that matters is not how much it answers — it is what it does when it does not know.Self-healing loopDocumentation doesn’t rot because people are lazy. It rots because nothing tells you it’s wrong.Two return paths nobody else has. A miss becomes a work order. A resolution becomes a draft. Usage improves the asset instead of wearing it out.