// Agent profile
WeKnora (Tencent/WeKnora) is an open-source, enterprise-grade knowledge platform built around three core modes: RAG-based Q&A for fast document lookups, a ReAct agent that autonomously orchestrates retrieval with MCP tools and sandboxed execution for multi-step reasoning tasks, and a Wiki mode where agents distill raw documents into a self-maintaining, interlinked markdown knowledge base with an interactive knowledge graph. It supports multi-tenant deployment, pluggable embeddings and rerankers, and is compatible with major LLM providers including OpenAI, Anthropic, DeepSeek, Qwen, Gemini, Ollama, and more. Written in Go, MIT licensed, 27K+ stars.
WeKnora is Tencent’s open-source knowledge platform, designed for teams that need to turn large document collections into something actually useful — not just a vector database with a chat interface bolted on. It organizes around three distinct modes, each solving a different knowledge work problem.
RAG Q&A handles everyday lookups: upload documents, ask questions, get sourced answers. The ReAct agent mode goes further — it autonomously plans multi-step research tasks, orchestrating retrieval, MCP tool calls, sandboxed code execution, and web search to answer complex questions that require reasoning across multiple sources. Wiki mode is the most distinctive feature: agents process your raw documents and produce a self-maintaining, interlinked markdown knowledge base complete with an interactive knowledge graph visualization.
ReAct agent with real tools: The agent mode is not a simple RAG chain. It uses ReAct-style reasoning to decompose complex questions, call MCP tools from a tenant-configurable skill catalog, execute code in session-persistent Docker or E2B sandboxes, and search the web — all while maintaining conversation context. This makes it capable of multi-step research and analysis tasks that pure retrieval systems cannot handle.
Wiki generation: Feed WeKnora a collection of raw documents — technical specs, research papers, internal docs — and its Wiki mode produces a structured, interlinked knowledge base. The knowledge graph visualization shows relationships between concepts, and the wiki updates itself as source documents change.
Multi-tenant enterprise deployment: WeKnora is built for organizations, not individuals. It supports multiple tenants with isolated document stores, configurable LLM backends per tenant, role-based access, and a skill catalog that administrators can customize per team.
Engineering teams use WeKnora to build searchable knowledge bases from scattered documentation, design docs, and runbooks. Research groups use the ReAct agent for literature synthesis and multi-source analysis. Organizations use Wiki mode to auto-generate and maintain internal knowledge bases that would otherwise require dedicated technical writers.
WeKnora is a platform-scale deployment — it runs as a Go service with dependencies on a vector database and your chosen LLM provider. It is not a lightweight CLI tool. The Wiki mode is powerful but requires tuning for domain-specific document structures. The project is primarily documented in Chinese and English, with the Chinese documentation being more comprehensive.
WeKnora is built for engineering teams, research organizations, and enterprises that need more than a basic RAG chatbot. It suits teams with large document collections who want autonomous research capabilities and auto-generated knowledge bases. If you need a multi-tenant, self-hosted knowledge platform with agentic capabilities, WeKnora is one of the most full-featured open-source options available.
AI agents that autonomously conduct ML research experiments on single-GPU training, running a keep-or-revert loop overnight.
Stealth headless browser for AI agents — wraps a Firefox fork with C++-level fingerprint spoofing to bypass Cloudflare, bot detection, and anti-scraping.
An AI-powered academic search engine that finds and synthesizes evidence-based answers from peer-reviewed scientific research.