# Univer: Open-Source Office SDK for Embedding Spreadsheets and AI Agents

> Source: <https://byteiota.com/univer-office-sdk-ai-agents/>
> Published: 2026-11-16 20:07:07+00:00

Univer, the open-source Office SDK from DreamNum, hit 988 new GitHub stars in a single day on November 16 — landing at 27,531 total and surfacing on GitHub Trending. This is not a new project breaking through; it is a mature one reaching critical mass. Developers are cloning it because it finally answers a question that has had no clean open-source solution: how do you embed a full spreadsheet, document, or presentation editor into your own web app without signing a commercial license agreement for Handsontable or SpreadJS?

## What You Get for Free

Univer 1.0 ships six editors under one Apache-2.0 license: Sheets, Docs, Slides, Boards, Bases, and PDFs. All six share the same plugin system, command system, and Facade API, so you are not stitching together six incompatible libraries. Installing it is straightforward:

```
npm install @univerjs/presets @univerjs/preset-sheets-core
```

No iframes. React-based view layer with support back to React 16.9. The open-source core includes 500-plus formulas, conditional formatting, data validation, permissions, comments, and hyperlinks. That covers most spreadsheet use cases for internal tools and product embeds.

Three things are behind the paid wall: real-time collaboration (which needs the Server SDK), XLSX and DOCX import/export, and charts plus pivot tables. The formula engine in the open-source tier is also slower than the commercial one. The [Business license starts at $99 per month](https://univer.ai/legal/license-agreement/). That is not nothing, but it compares favorably to per-developer commercial grid pricing.

## The Architecture That Sets It Apart

The more interesting story is what Univer is built for. The team positioned it explicitly as an AI agent platform — and they did not mean “we added a chatbot to the sidebar.” The design assumption is that agents and humans will be editing the same documents.

Univer handles this with a Worktree pattern borrowed from git. Agents work in isolated draft copies of a document, not the live file. When an agent finishes its changes, a human sees a side-by-side semantic diff and decides: approve, reject, merge, or roll back. This applies to sheets, docs, and slides equally.

The [AI SDK](https://docs.univer.ai/ai) gives agents four tools to work with: `univer_new` to create files, `univer_execute` to read and edit content via the Facade API, `univer_screenshot` to render the current state as a PNG for visual verification, and `univer_api` to look up available API symbols. The screenshot tool in particular is worth noting — it lets an agent verify what a document actually looks like after edits, not just what the data model says.

## If You Are Still on Luckysheet

DreamNum built Luckysheet before Univer. Luckysheet was widely adopted, MIT-licensed, and functional. It is now in maintenance-only mode. The same team moved their attention to Univer entirely, and there is no indication Luckysheet will receive new features.

The migration is not a drop-in replacement — the APIs differ substantially. If you have a Luckysheet dependency in production, the question is not whether to migrate but when. Univer is the successor the same team is developing, and the gap between them will only widen.

## Who This Is For

Univer makes sense for three scenarios. First: teams building internal tools or SaaS products that need embedded spreadsheet or document editing and want Apache-2.0 licensing. Second: teams that need agents to operate office documents in a reviewable, controlled way — the Worktree pattern is a real architectural advantage over manually managing agent edits. Third: developers on Luckysheet looking for an upgrade path.

It does not make sense for data table display. If you are showing rows of records with sorting and filtering, [TanStack Table](https://github.com/TanStack/table) or AG Grid are better fits. Univer is a spreadsheet, not a data grid — the distinction matters when you are choosing a library.

The [GitHub repository is at github.com/dream-num/univer](https://github.com/dream-num/univer). [Installation docs](https://docs.univer.ai/guides/sheets/getting-started/installation) and the [AI SDK reference](https://docs.univer.ai/ai) are both on docs.univer.ai. The [Univer Workspace](https://github.com/dream-num/univer-workspace) repo is worth bookmarking if you plan to self-host a full collaboration layer.
