WeKnora is Tencent's open-source, LLM-powered knowledge framework for enterprise document understanding, semantic retrieval and reasoning, written in Go. It works in three modes over the same knowledge bases: RAG Q&A that answers from the documents and cites its sources, a ReAct-style agent that plans multi-step work using knowledge-base search, web search, MCP tools, skills and the user's local browser, and an optional Wiki mode that turns documents into interlinked, source-cited pages with a knowledge graph, in-browser editing and rollback. Retrieval combines keyword and vector hybrid search with reranking, parent-child chunking and GraphRAG through Neo4j, and the project includes end-to-end evaluation with recall and BLEU/ROUGE metrics. Agent skills can be installed from ClawHub, SkillHub, Git or ZIP and run in session-persistent Docker, E2B or Cube sandboxes with network policy. Release v0.8.2 adds the BrowserSkill local-browser extension and a built-in MCP server over Streamable HTTP, and notes two breaking changes: DingTalk channels are Stream-only and sandbox commands run as root. Parsing covers PDF, Word, PPT, Excel, EPUB, XMind, images and more, with data-source sync from Feishu, Confluence, GitLab, Notion, Yuque, DingTalk Docs and RSS. It supports 27 built-in model vendors including OpenAI, Anthropic, DeepSeek, Qwen, Gemini and Ollama, several vector databases such as pgvector, Elasticsearch, Milvus and Qdrant, workspace RBAC with an audit log, and Langfuse tracing. Deployment options are Docker Compose, Helm, a Lite single binary and a desktop app that has no published installer yet and must be built from source. The README recommends running it on an internal network rather than exposing it to the public internet. The code is MIT-licensed, with listed third-party components under their own licenses.