kbaze

kbaze is a wiki platform designed for teams that includes built-in semantic search and artificial intelligence capabilities. It allows users to write pages in Markdown, link them into a knowledge graph, and query their documentation through an AI chat that cites its answers. The product supports Mermaid diagrams, spaces with role-based permissions, and version history with authorship and commit messages.

Access to kbaze at the time of the scraped text was limited to early access by invitation only. Users can enter by accepting an invitation and logging in with their Google account to create a workspace for their team. There is no dedicated administrator required to get started.

kbaze is built serverless on AWS, meaning it operates without servers that need to be maintained. Documentation is stored in isolated, encrypted workspaces, and the infrastructure is described as serverless on AWS. This architecture ensures that data is kept secure and separate between different teams or projects.

The platform includes AI as a core feature rather than charging extra for it. The same semantic index that powers the search also feeds the assistant, allowing for chat about documentation with cited answers. This means the AI capability is included in all spaces without needing a separate plan.

kbaze supports the MCP protocol to connect external agents, such as Claude Code. These agents can consult and write in the wiki with credentials that are scoped by space and can be revoked instantly. This integration allows for automated workflows while maintaining strict control over access.

Users can create diagrams using Mermaid or draw freely as editable blocks rather than pasting images. The editor supports both WYSIWYG and Markdown, ensuring no format loss during conversion. These features allow teams to visualize processes directly within the documentation.

The platform offers role-based permissions ranging from owner to reader, along with groups and draft modes. Each user sees only what corresponds to their role, and explicit publication controls manage drafts. This structure ensures that sensitive information remains accessible only to the intended audience.

Every publication is versioned with author and comment details, allowing users to compare, restore, and audit changes. This version history does not require additional plugins and tracks who changed what. It provides a clear audit trail for all modifications made to the documentation.

The search functionality is hybrid, combining lexical searches for exact matches with semantic searches for meaning. For example, it can find a page titled 'runbook' when searching for 'how to deploy to production'. This dual approach helps users locate relevant content even if the exact keywords are not used in the title.

kbaze is designed for teams looking to create a single source of truth where all knowledge is connected. It suits organizations that want to move away from scattered folders and chat channels to a unified knowledge graph. The platform is particularly useful for teams that need to query their documentation with AI assistance.

kbaze allows users to export their content as Markdown, ensuring that their documentation remains portable and open. The platform emphasizes that the content belongs to the user, with open export formats. This feature supports long-term data ownership and flexibility in moving between tools.

The platform includes a feature where the AI assistant respects permissions, only consulting what the user can see. Private drafts remain outside the assistant's scope, ensuring confidentiality. This design prevents the AI from accessing or summarizing information that is not visible to the current user.

The official website of kbaze is kbaze.com.