Your AI can write an email or summarize a meeting transcript. Ask it to pull numbers from a spreadsheet, update a proposal document, and export the result as a PDF in a single go, and it usually falls apart somewhere in the middle. That just got a lot easier.
What's happening
Univer 1.0 just landed. It is a free, open-source office toolkit (spreadsheets, documents, slides, boards, databases, and PDFs) that is built from the start to let AI agents work inside it, not just read files out of it. Apache 2.0 license. Runs on your own server or laptop. No subscription.
What makes 1.0 the milestone worth noticing: it ships a native AI interface that lets an agent move fluidly across all six document types in a single workflow. Your agent can pull from a spreadsheet, write findings into a document, and push that to a PDF, without you babysitting the handoffs between three separate tools. The team also rebuilt Excel and Word compatibility for this release, so your existing .xlsx and .docx files come in with formulas, charts, and styles intact, not just stripped plain text.
This is not another "AI can summarize your doc" wrapper. It is a layer that lets a properly set-up AI agent actually do work in your documents, the way a contractor works in a shared drive, except none of your data leaves your building.
Try this this week
- Go to docs.univer.ai and run the five-minute quickstart to spin up a local instance on your machine
- Try a simple cross-type test: put a small table of numbers in a Sheets tab, then ask your AI tool to read it and write a one-paragraph summary into a Docs page
- If your business runs on Excel files, drop one in and confirm your formulas and charts survive the import (the 1.0 release specifically fixed this compatibility)
- Check the AI SDK examples in the docs if you want to wire an agent to a live document directly
Works on Mac, Windows, and Linux. You can point multiple AI agents at the same workspace at the same time.
The bigger picture
What used to mean stitching together a document API, a spreadsheet API, and a PDF library now ships as one free package you run on your own terms. The Univer project is small team, open source, and not chasing enterprise contracts. That is exactly the kind of tool build9 watches for: something that used to cost real money, or require real DevOps work, that a five-person shop can now stand up in an afternoon. The smallest working thing is the one you control from day one.