WeKnora

WeKnora is a self-hosted AI knowledge platform for teams that want to ask questions about their documents, use them in agent tasks and organize them into a wiki. These functions share the same knowledge base. It supports local inference through Ollama and offline deployments, with replaceable model, database and storage backends.

Document Q&A combines keyword and semantic search, then attaches source citations so readers can check answers against the original material. It accepts PDFs, Office files and images, and can sync content from Confluence, GitLab, Notion, Feishu and other services. The wiki turns documents into linked pages with citations and a knowledge graph. Teams can edit pages, compare revisions and restore earlier versions.

The agent can search the knowledge base and web, call MCP tools, run skills and create files in a sandbox. A terminal and graphical desktop let users inspect its work or take over. BrowserSkill lets it operate the user's own Chrome or Edge, handing control back for logins and CAPTCHAs. Conversations support branching and rollback, and the agent can retain preferences users have confirmed across sessions.

Teams can run WeKnora in Docker or Kubernetes, use its single-machine edition or desktop app, or deploy it on a private cloud. A separate hosted service provides knowledge Q&A for WeChat. Local deployments can keep data in the team's environment; cloud model providers are optional. Workspace permissions, OIDC and audit logs support team access, while the built-in MCP server makes knowledge bases available to Claude and Cursor.