WeKnora
An open-source, LLM-powered knowledge platform that turns a team's documents into a queryable retrieval-augmented system, a multi-step reasoning agent, and a self-maintaining wiki, all sharing the same knowledge bases.
WeKnora indexes documents from sources such as Feishu, Confluence, GitLab, Notion, Yuque and RSS feeds, including PDF, Word and Excel formats parsed in-process, and lets a team look things up with retrieval-augmented search, hand multi-step tasks to an agent, or maintain a wiki, all against the same underlying knowledge bases. The agent's tools -- skills installed from ClawHub, SkillHub, Git or a ZIP file -- run in session-persistent Docker, E2B or Cube sandboxes with an interactive terminal and desktop alongside the chat, and it can operate the user's own Chrome or Edge browser or connect external MCP services.
Other features include cross-session long-term memory of confirmed facts and preferences, editable and rollback-able retrieval chunks, Q&A integrations for WeCom, Feishu, Slack and Telegram, an embeddable widget, a built-in MCP server for tools such as Cursor and Claude, multi-workspace role-based access control with per-workspace audit logs, and Langfuse tracing of agent steps and token usage. It supports more than 25 model vendors and swappable vector database and storage backends, and can be deployed locally, on a private cloud, or through a managed WeChat integration.
It is written in Go.