Building an open source knowledge base for the AI era
We have seen the need with our clients, and we have it ourselves. At drunomics we work with AI agents every day. They write code, draft documents and answer questions from our internal docs. And agents are only as good as the knowledge they are given. Outdated knowledge is a real problem here: an agent reads a stale runbook, gets distracted by a decision that was reverted months ago, and confidently does the wrong thing. A person would ask a colleague. An agent trusts the page and goes the wrong way.
So a tidy, up-to-date knowledge base is no longer a nice-to-have. It is the foundation every agent, assistant and workflow stands on. And it works both ways. When a teammate shares a page with you, you need to know it is not just some random AI-generated content, but actually verified content you can rely on as if they had written it themselves. Telling an AI-generated draft apart from carefully maintained, reviewed and corrected knowledge is crucial.
The pieces are already there
Looking at what we build with every day, we figured the tools for a great knowledge base already exist, and all of them are open source! There is Drupal's robust backend and the Drupal AI tooling, the Drupal MCP server, Lupus Decoupled, Nuxt building blocks like comark (component-enhanced Markdown, the ideal format for AI), Nuxt Chat and the Nuxt UI Editor, Hocuspocus for collaborative editing, and OpenSearch for search and vectors.
Put together right, that is a knowledge base built for the AI era: agents as first-class citizens, working alongside people in the same collaborative session. With trust built in: configurable review levels for agents and people per space, backed by Drupal's powerful governance tools for permissions, workflows and revisions, plus provenance and citations down to the block. As long-time Drupal contributors and Drupal AI Silver Maker, we worked it out with the Drupal AI initiative to build this in collaboration. So it becomes a fully working product that shows what Drupal AI can do. At the start that means the foundation, the built-in chat assistant, RAG with citations and AI-enhanced search. From there we plan to take advantage of more sophisticated Drupal AI features, like background agents that help you keep your knowledge up to date and free of conflicts.
Introducing OpenKnowledgebase
So we are happy to introduce OpenKnowledgebase - the open source wiki & AI knowledge base for people and agents.
We are excited to finally share what we have been building: OpenKnowledgebase, the open source wiki & AI knowledge base for people and agents. Four things carry it:
- Agents as governed collaborators. Named, permissioned, reviewed before their changes count.
- Knowledge you can trust, down to the block. Provenance and citations per block, not per page.
- Real collaboration. Live co-editing, inline comments, spaces, Markdown underneath.
- Open and sovereign. MIT and GPLv2+, no open-core. Your data, your models, your code, wherever you run it.
We develop it fully as open source, not open-core. The self-hosted edition is the whole product, under MIT and GPLv2+. Around it, we plan to offer services for those who want to save time or need more: managed hosting as a SaaS, varying support levels and migration services. That makes it a viable business model that funds further development of the open source product, without withholding features.
Beta 1 is coming
We have been working on this for the last eleven weeks, and we are getting the first beta ready. Expect the first public beta in the upcoming weeks. Until then: openkb.app has the overview, and the newsletter gets you one mail when the beta is out. If you are at DrupalCon Rotterdam, come and talk to us - we would love to hear what you think!
