One knowledge base your team and your AI tools both trust
Build an internal knowledge base for SOPs, runbooks, and onboarding, with role-based access, AI search, and automations that flag content the moment it goes stale.
Live in under 5 min · No credit card required
Trusted by product teams at
Internal wikis fail in the same four ways
Nothing is findable
Search matches words rather than intent, and the document you need was titled by somebody who left last year.
Nothing is current
There is no signal for staleness, so every page looks equally authoritative and a good share of them are wrong.
Access is all or nothing
Either everyone sees everything, or the useful half sits in a space nobody has thought to request access to.
AI tools cannot reach it
The knowledge exists, and the assistant your team now asks first has no way to read a word of it.
Private by default, structured for retrieval
- Private mode requires authentication on every page; partial mode gates only the pages you choose
- Password, JWT, or OAuth 2.0 with Okta, Auth0, Azure AD, or any OIDC provider, with per-user roles
- Hybrid semantic and keyword search, plus an AI Assistant that answers with citations
- Structured components give models the semantic cues that make retrieval accurate rather than approximate
- An authenticated MCP endpoint so Claude, ChatGPT, and Cursor search gated content under your permissions
- Scheduled automations for links, grammar, style, and navigation, each returning a reviewable commit
Roles carry through everywhere. An MCP client gets exactly the access its key allows, and a gated page stays gated in search, in the assistant, and over MCP.
The three things a wiki never had
Access control on every surface
Three access modes, three authentication methods, and per-user roles on JWT and OAuth. The same permissions apply in the site, in search, in the AI Assistant, and over MCP.
Retrieval built into the content model
Structured components are not decoration. They give models the semantic cues that keep chunking clean, which is the difference between an assistant that quotes your runbook and one that paraphrases it wrong.
Staleness that surfaces itself
Scheduled audits catch broken links and navigation problems, while feedback and low-confidence assistant answers point at the pages your team tried to use and could not.
The answer, without interrupting a colleague
- Works on private and partial sites, under the permissions you already set
- Citations point at the runbook, so the reader can follow the full procedure
- The same rules apply over MCP, for the AI tools your team runs
Built for content that cannot be public
3 access modes
Public, partial, and private, set per project.
Per-user roles
JWT and OAuth 2.0 carry access roles and support personalisation.
Authenticated MCP
OAuth 2.1 for interactive clients, machine credentials for automation.
Cited answers
The assistant links the page and section behind every answer it gives.
Frequently asked questions
What teams verify before internal knowledge moves in.
Can we use our existing SSO?
Yes. OAuth 2.0 and OpenID Connect providers including Okta, Auth0, and Azure AD, or JWT tokens issued by your own application.
Can different teams see different pages?
Yes. JWT and OAuth support per-user access roles, and partial mode lets you mark individual pages as public or protected.
How does an internal knowledge base stay current?
Automations run on a schedule or after a pull request merges, covering broken links, grammar, style guide compliance, and navigation. Each run returns a summary, a report, and a commit to review.
Could an AI assistant leak a private page?
No. Access-controlled sites use an authenticated MCP endpoint, and a client reads only what its own credentials allow.
Ready to build documentation people and AI agents love?
Live in under 5 min · No credit card required