Upload an Excel or CSV file and let Kimi clean the data, find trends, write formulas and build pivot tables and charts, then download a finished .xlsx file. Here's how, with prompts, formula examples and ways to check the numbers.
Use chat to ask about a file, and Kimi Sheets when you want a finished Excel workbook back.
| What you need | Use | What you get |
|---|---|---|
| Quick questions about a file | Kimi chat | Attach the file and ask. Answers come back as text and tables in the chat. |
| A finished .xlsx with formulas, pivots and charts | Kimi Sheets (Agent) | Builds a real Excel file you can preview and download. Uses agent credits. |
| Files and folders on your computer | Kimi Work | The desktop app works on local files, including an Excel plugin for spreadsheet tasks. |
| Many files in a script or app | Kimi API | Upload files with the Files API and process them in code. |
More on the desktop app: Kimi Work guide. Credits and plans: Kimi pricing.
Sign in at kimi.com, then choose Sheets in the sidebar or click the Agent button under the message box.
Attach your Excel (.xlsx) or CSV file. Each file can be up to 100 MB, and you can add up to 50 files.
Say what to calculate, how to group it and what to output, for example “Revenue by region and month as a pivot table, plus a bar chart.”
Kimi shows the sheets it built. Check the structure, then ask for changes in plain words.
Download the .xlsx, open it in Excel, Google Sheets or WPS, and check the key totals and formulas yourself.
Analyse the attached spreadsheet. Explain what the data shows, give the 5 most important findings with the numbers behind them, and build a summary sheet with a pivot table and one chart. Use formulas, not typed values, and flag any data problems.
New to Kimi? Read the beginner's guide first.
Attach your file, copy a prompt and replace the [brackets] with your columns and goals.
Clean the attached spreadsheet: remove duplicate rows, set all dates to YYYY-MM-DD, trim spaces, make amounts numeric and standardise names and country codes. Don't guess missing values; flag them. Add a sheet called 'Cleaning log' listing every change.
Analyse the attached sales data. Describe what each column contains, then give me the 5 most important trends, the best and worst performers, and anything unusual. Show the numbers behind each point.
Create a pivot table from the attached data showing total [revenue] by [region] (rows) and [month] (columns), with grand totals and each region's share of the total as a percentage.
Build a dashboard sheet from the attached data: 4 key numbers at the top, a line chart of [revenue] by month, a bar chart by [product], and a top-10 table. Keep the charts linked to the data.
I need a formula in [Excel] that [goal, e.g. totals sales for the North region in March]. My columns are: [A: Date, B: Region, C: Product, D: Revenue]. Give the formula, explain it step by step and show one worked example.
Combine the attached [12] monthly files into one sheet. Match columns by name, add a 'Source file' column, remove exact duplicates, and add a summary sheet with monthly totals.
Check the attached spreadsheet for problems: totals that don't add up, outliers, negative values where they shouldn't be, missing data and duplicates. List each issue with its row number and how sure you are.
Using the monthly data in the attached file, forecast the next [6] months of [revenue]. Show low, base and high scenarios, explain the method in plain words, and keep every assumption in its own cell so I can change it.
Compare the budget and actual figures in the attached file. Add variance and variance % columns, highlight lines over budget by more than [10%], and write a 5-line summary for my manager.
More ideas in the Kimi prompt library.
Most wrong answers come from messy data: mixed date formats, numbers stored as text, extra spaces and the same customer spelled three ways. Ask Kimi to clean the file before you analyse it.
Messy rows become a consistent table, with problems flagged instead of guessed.
Describe what you want and your column layout. These examples assume Date in column A, Region in B, Product in C and Revenue in D.
| You ask Kimi for… | Formula it can write | Copy |
|---|---|---|
| Total sales for one region | =SUMIFS(D:D, B:B, "North") | |
| Total for one region in March 2026 | =SUMIFS(D:D, B:B, "North", A:A, ">=2026-03-01", A:A, "<=2026-03-31") | |
| Count orders over $1,000 | =COUNTIFS(D:D, ">1000") | |
| Look up a price by product code | =XLOOKUP(F2, Products!A:A, Products!C:C, "Not found") | |
| Month-on-month growth | =IFERROR((C3-C2)/C2, "") | |
| Running total | =SUM($D$2:D2) | |
| Flag duplicates | =IF(COUNTIF(A:A, A2)>1, "Duplicate", "") |
Tip: add “explain it step by step” to learn the formula, and say whether you use Excel, Google Sheets or WPS, since a few functions differ.
Kimi is strong at spreadsheet work, but it can still miss rows or a mismatch. In Puter's October 2026 finance test, K3.1 Pro didn't make up any figures, but it missed a $20M gap that Kimi K3 pointed out in two of three runs. Reconcile the numbers yourself.
Compare Kimi's grand totals with a quick SUM in your own sheet. If they don't match, ask Kimi which rows it left out.
Click on a few result cells. Live formulas update when data changes; typed-in numbers don't. Ask Kimi to “use formulas, not hard-coded values”.
Pick three rows and work them out by hand. It takes two minutes and catches most errors.
Tell Kimi to flag missing or odd values instead of filling them in, and to say which ones it changed.
Moonshot built K3.1 Pro for finance work, from research to finished workbooks with linked formulas. These are Moonshot's own reported results.
Analysis of real financial documents; the top score in Moonshot's comparison.
Points gained over K3 on formula linkage, financial logic and workbook quality.
Points gained over K3 on research that has to hold up to a fact check.
K3's score on a long simulated business-management task.
| Limit | What it is | Source |
|---|---|---|
| File types | Excel and CSV, plus PDF, Word, PPT, images and TXT as sources | Kimi Help Center |
| File size | Up to 100 MB per file | Kimi Help Center |
| Files at once | Up to 50 files | Kimi Help Center |
| Row limit | Not published | Very large files may need splitting or summarising first |
| Output | .xlsx you preview online, then download; Excel, CSV and other formats | Kimi Help Center |
| Cost | Chat questions are free within limits; Sheets runs as an agent task and uses credits | Kimi plans |
Working from PDFs instead? Kimi Sheets can pull tables out of PDFs into Excel. See how to analyze PDFs.
For many files or a repeating report, upload CSV or Excel files with the Files API, or summarise large data in pandas first and send only the summary.
purpose="file-extract" and send the extracted text as a system message.K3.1 Pro costs $2 per million input tokens and $10 per million output tokens. Keys and setup: Kimi API guide.
Other issues? See Kimi troubleshooting.