WeKnora is a free, MIT-licensed tool that turns a team's existing documents into a self-updating internal wiki, a job many companies currently pay Confluence or a similar tool to do.
A support team that has ever tried to keep a Confluence space current knows the real cost isn't the license. It's the hours someone spends re-reading old pages, merging duplicates, and figuring out which version of the onboarding doc is actually correct. WeKnora, an open-source project from Tencent, tries to remove that maintenance step. Point it at a folder of PDFs, Word files, spreadsheets, or scanned documents, and it builds a searchable, interlinked wiki from them automatically, then keeps rewriting that wiki as the source documents change. Confluence's Premium tier, the plan most growing teams eventually need for unlimited storage and real admin controls, runs $10.44 per user per month. WeKnora is free to run and MIT licensed.
The project describes itself as three things at once: a RAG-based question-answering layer over a company's documents, an agent that can chain together retrieval, web search, and other tools for multi-step questions, and a "Wiki Mode" that distills raw uploads into structured markdown pages with an interactive knowledge graph. That last piece is the part worth paying attention to. Rather than a static export, the generated wiki carries page-level revision history and one-click rollback, so an editor can correct or revert a page the same way they would in any real wiki tool, while the underlying documents keep feeding updates back in.
It handles more than a dozen document formats, including PDF, Word, Excel, PowerPoint, and XMind files, and it can sync automatically from Feishu, Notion, Yuque, GitLab, and a handful of other sources rather than requiring a one-time upload. Teams can also expose the resulting Q&A through Slack, WeCom, Telegram, or a handful of other chat apps, so the wiki isn't just a page someone has to remember to visit. The most recent tagged release, v0.8.0, shipped a sandboxed "skill" runtime and cross-session memory on top of the existing retrieval and wiki features, on a repository with 25,100 stars and 3,500 forks and an active commit history through this week.
For a mid-sized company, the appeal is straightforward: internal knowledge tends to live scattered across drives, chat threads, and a wiki that only gets updated in bursts. A tool that regenerates that wiki from whatever documents already exist removes a task that usually falls to whoever has the least time for it. That makes WeKnora more interesting to an internal IT or ops function than to an individual marketer trying to prototype something over a weekend. Standing it up means running Docker Compose, wiring up a Postgres-backed vector store, and connecting an LLM provider, whether a paid API like OpenAI or a locally hosted model through Ollama. None of that is difficult for a team with any development capacity, but it isn't a five-minute signup either.
The honest tradeoff is that "free" here means free of a subscription, not free of work. Someone still has to run the containers, keep the deployment patched, and cover whatever LLM inference costs the wiki generation and Q&A actually incur, which scale with usage in a way a flat per-seat Confluence bill does not. The project's own security guidance recommends deploying it on an internal network rather than exposing it to the public internet, which is a reasonable default but one more thing for whoever owns the deployment to get right. And a business that leans on Atlassian Guard, granular admin controls, or a 99.9% uptime SLA on the Premium plan is trading a supported product for a self-managed one.
What WeKnora points at, more than the tool itself, is a shift in what a "cheap" internal system actually saves. It used to be that self-hosting saved you a subscription and cost you setup time. Now the setup is mostly solved, and the tradeoff moves to who is accountable for a document a model wrote and nobody double-checked.