Tencent Turns RAG Knowledge Base into an Agent: WeKnora Can Now Remember, Call Tools, and Even Run Code

Beating AI News Flash: Tencent's WeChat team continues to expand the open-source knowledge framework WeKnora. It originally focused on turning enterprise documents into searchable, Q&A-capable RAG knowledge bases, and later gradually added Agent, MCP, and Skills. The latest v0.8.0 now adds cross-session long-term memory and further completes Skills and the code execution environment.

WeKnora can now remember users' materials, preferences, facts, and tasks, and continue to use them in the next chat. Memories automatically extracted by the system do not take effect directly and must first be confirmed by the user. Skills have also been made into independent workspace resources and can be installed from ClawHub, SkillHub, GitHub, GitLab, or zip files.

When executing tasks, the Agent can also enter Docker, E2B, or Cube sandboxes, run commands, read and write files, and then return generated files directly to the chat window. Administrators can restrict which networks the sandbox can access. WeKnora's focus is still enterprise knowledge, but it is now no longer just "finding answers from documents"; it has begun to put knowledge, memory, tools, and the execution environment into the same Agent system.