Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/manusapis/Agix/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks you from a fresh checkout to a working Agix sidebar in Excel Desktop. By the end you will have the task pane open, an API key saved, and the AI responding to a question about a cell range you select. The whole process takes about five minutes.
1

Install prerequisites

Agix requires Node.js 18 or later. Verify your version before continuing:
node --version   # must be v18.0.0 or higher
You also need a copy of Excel 2016 or later (Desktop or Excel on the web) to sideload the add-in. Excel is available through Microsoft 365 on Windows and Mac.
2

Clone the repository

Clone the Agix repository and move into the project directory:
git clone https://github.com/manusapis/Agix.git
cd Agix
3

Install dependencies

Install all Node packages with npm:
npm install
This pulls in Office.js, React 18, Vite, and TypeScript — roughly 400 MB of node_modules on a first install.
4

Start the development server

Start the Vite dev server, which serves the add-in over HTTPS on port 3000:
npm run dev
You should see output similar to:
VITE v5.x  ready in 300 ms

➜  Local:   https://localhost:3000/
➜  Network: https://192.168.x.x:3000/
Leave this terminal running. Excel loads the task pane from this local server.
Excel requires a trusted HTTPS certificate to load Office Add-ins from localhost. If your browser shows a certificate warning when you visit https://localhost:3000, install a locally-trusted certificate with mkcert before sideloading:
mkcert -install
mkcert localhost
Then configure Vite to use the generated localhost.pem and localhost-key.pem files. Without a trusted cert, Excel will silently fail to load the task pane.
5

Sideload manifest.xml in Excel

Tell Excel about the add-in by uploading the manifest file:
  1. Open Excel (Desktop).
  2. Go to Insert → My Add-ins.
  3. Click Manage My Add-ins → Upload My Add-in.
  4. Browse to manifest.xml in the project root and select it.
  5. Click Upload.
Excel registers the add-in and adds an Agix AI group to the Home ribbon.
<!-- manifest.xml excerpt — the SourceLocation that Excel loads -->
<SourceLocation DefaultValue="https://localhost:3000/taskpane.html" />
The manifest points to https://localhost:3000, which is why the dev server must be running first.
6

Open the Agix task pane

In the Home ribbon, click Open Agix inside the Agix AI group. The sidebar slides in on the right side of the workbook.You will see:
  • A Provider dropdown (defaults to OpenAI)
  • A chat textarea for your prompt
  • A Send button and a Use selection button
7

Enter your API key

Agix does not bundle any API keys. You need to supply a key from your chosen provider before sending your first message.
  1. In the sidebar, locate the Provider dropdown and select the provider you want to use (OpenAI, Claude, or OpenRouter).
  2. Enter your API key in the API key field.
  3. The key is saved automatically to your user settings and persists across sessions.
If you do not have a key yet, create one at:
8

Select a range and send your first prompt

  1. Click on any range of cells in your worksheet that contains data — for example a column of sales figures.
  2. In the Agix sidebar, click Use selection. The active range address loads into the prompt field.
  3. Edit the prompt to ask a real question, such as:
Sheet1!A1:B12 — What is the month-over-month growth rate for each row?
  1. Click Send. Agix forwards the range reference and your question to the AI provider, then streams the response into the chat panel below.
The Use selection button reads the currently highlighted cells in Excel and pastes the range address into the prompt. You can then type your question after the address, or clear the field and describe the range in your own words. This is the fastest way to ground a question in real sheet data.

​
Other useful commands

Once the project is set up, you can run type checking and linting from the project root at any time: 
npm run typecheck   # runs tsc --noEmit; reports TypeScript errors without emitting files
npm run lint        # runs ESLint across src/ for .ts and .tsx files
npm run build       # production build; outputs static assets to dist/

​
What’s next

AI Providers

Switch from the default GPT-4o mini to Claude, OpenRouter, or a custom endpoint — and learn how API keys are stored securely.

Chat features

Learn how to use the chat panel to generate formulas, analyze ranges, ask follow-up questions, and write results back to cells.

Build docs developers (and LLMs) love

Get started for freeTalk to us