An internal knowledge base that tells you when it's going stale.
An AI knowledge base for your runbooks and docs: as many as you need, each with a purpose, required sections and a review queue. Documents are stored as a graph with hybrid search. The assistant answers across tickets, issues and docs in the app, and over MCP the knowledge base is the RAG system behind Claude, ChatGPT and your other AI tools. A health score tells you what is stale, orphaned or missing.
How it works
From a blank page to an internal knowledge base that stays healthy.
Set it up
Give each knowledge base a purpose and the sections every page must answer. Grant read, write or full access per team.
Write
New pages are checked against the profile. Pages that fail are held in the review queue, out of search, until someone approves them.
Connect
Documents and the links between them are stored as a graph with embeddings, so search works by relationship and by meaning at once.
Ask
The assistant answers across knowledge bases, tickets and issues, in the app or from Claude, ChatGPT, Copilot and Gemini CLI over MCP.
Keep it healthy
Knowledge Health flags stale pages people still open, orphans and searches that found nothing. Recompute it whenever you like.
What you get
Everything in Knowledge Base.
Graph storage
Documents and the links between them are stored as a graph, with embeddings for meaning.
Hybrid search
Search by relationships and by meaning in the same query.
Graph view
Pan, zoom and click to highlight connections, with a reading panel on the side.
AI assistant
Ask about your tickets, issues and knowledge bases. Every chat is saved, so you can pick it up again later.
Profiles and required sections
Each knowledge base has a purpose and the headings every page must answer. A section that cannot be answered honestly is one a page cannot fake.
Review queue
New pages that fail an automated check are held for review and kept out of search until someone approves them. Hold every page if you prefer.
Access by team
Read, write or full access per team and per knowledge base. Owners see everything.
MCP connectors
One connector per member for Claude, ChatGPT, GitHub Copilot or Gemini CLI. It signs in as that member and reaches their knowledge bases, tickets and work log.
RAG for your AI tools
Use any knowledge base as the RAG system for Claude, ChatGPT, Copilot, Gemini CLI or another AI tool that supports MCP. It retrieves with the same hybrid search as the app, within each member's access.
Knowledge Health
Flags stale documents people still open, orphans and unmet demand in search. Recompute it whenever you like.
Activity log
Create, update, rename and delete per member, plus an audit of every assistant action, in the app and over MCP.
Soft delete
Owners and admins can bring a deleted knowledge base back.
In depth
A closer look.
Profiles and review
Pages that answer the questions you decided matter.
Most knowledge base software takes any page you give it. Here each knowledge base has a purpose and the headings every page must answer. A runbook without a rollback section, or an owner left blank, fails the check and is held in the review queue, out of search, until someone approves it. A section that cannot be answered honestly is one a page cannot fake.
- Required sections per knowledge base
- An automated check on every new page
- The review queue keeps unapproved pages out of search
- Hold every new page if you prefer
RAG over MCP
Your knowledge base as the RAG system behind your AI tools.
Connect Claude Code, Claude, ChatGPT, GitHub Copilot, Gemini CLI or another AI tool that supports MCP, and it retrieves from your documents, tickets and work log as it answers, with the same graph and vector search as the app. There is no pipeline to build: the app chunks, embeds and links your documents itself. Each member's connector signs in as them, so it reaches only what they can see, and every action is in the activity log.
- Claude, ChatGPT, GitHub Copilot and Gemini CLI
- No chunking, embeddings or vector database to run
- Retrieval within each member's access
- Pages held for review stay out of retrieval
Knowledge Health
Know which page to fix first.
Each knowledge base gets a health score you can recompute whenever you like. It lists the old pages people still read, ranked by reads, the searches that found nothing, pages that overlap or disagree, orphans nothing links to and anything that looks like a secret. Answer quality shows how people rate the answers each document gives, and every finding has a button that asks the assistant to draft the fix.
- Stale pages still in demand, ranked by reads
- Searches that found nothing, ranked by demand
- Duplicates and pages that disagree
- Retrievals, ratings and match quality per document
Connects to
The assistants your team already uses.
Who uses it
An AI knowledge base for the people who have to trust it.
Questions
Knowledge Base FAQ.
How many knowledge bases can we have?
As many as you need. Each has its own purpose, required sections, review settings and team access, and owners see all of them.
What does the review queue hold back?
New pages that fail the automated check against the knowledge base's profile, such as a required section left empty. They stay out of search until someone approves them. You can also hold every new page if you prefer.
What does Knowledge Health measure?
Three things: stale documents people still open, orphaned documents nothing links to, and searches that found nothing. Together they make a score you can recompute whenever you like.
Which AI assistants can connect?
Claude, ChatGPT, GitHub Copilot and Gemini CLI, through one MCP connector per member. The connector signs in as that member and reaches their knowledge bases, tickets and work log with the same access checks as the web app.
Can Claude Code read our internal docs?
Yes. Every knowledge base is also an MCP knowledge base: each member gets an MCP connector that works in Claude Code as well as Claude, ChatGPT, GitHub Copilot and Gemini CLI. It signs in as that member and searches the knowledge bases they can see, plus their tickets and work log.
Can we use the knowledge base as a RAG system for our AI tools?
Yes. Every knowledge base is also a RAG knowledge base over MCP. Connect Claude, Claude Code, ChatGPT, GitHub Copilot, Gemini CLI or another AI tool that supports MCP, and it retrieves from your documents, tickets and work log with the same hybrid search as the app. The app chunks, embeds and links your documents, so there is no pipeline or vector database to run, and each connector reaches only what its member can see.
What is the difference between RAG and MCP?
RAG, retrieval augmented generation, means a model looks up your own content before it answers. MCP, the Model Context Protocol, is the open standard an AI tool uses to reach data and tools. Here they work together: MCP connects your AI tool to the knowledge base, and the knowledge base does the retrieval, so you get RAG without building it.
How does search find the right page?
Documents and the links between them are stored as a graph, with embeddings for meaning, so one query searches by relationship and by meaning at once: graph and vector search together. Your AI tools get the same retrieval over MCP, graph RAG and vector RAG in one.
Is the assistant's activity logged?
Yes. Every assistant action, in the app and over MCP, is in the activity log next to who created, updated, renamed or deleted a page.
What if someone deletes a knowledge base?
It is soft deleted. Owners and admins can bring it back.
How is Knowledge Base priced?
Per seat for each person who can reach Knowledge Base, plus storage per gigabyte. The pricing page has the numbers.
Part of one workspace