v0.6.0 is out · Data quality checks, dashboards, editable query results

Your database,
finally friendly.

Kiyi lets anyone browse, search and edit PostgreSQL, MySQL, MariaDB, SQLite and SQL Server like a spreadsheet, and ask for data in plain words. No SQL required. All of it there the moment you want it.

macOS, Windows and Linux · MIT licensed · No account needed

Kiyi showing a customers table with a row open in the details panel
Saved Undo
“pro customers in germany who spent over 1k”
SSH tunnel · Signed in
Works with PostgreSQLMySQLMariaDBSQLiteSQL Server
Bring your own AI AnthropicOpenAIGeminiOpenRouterxAI GrokGroqMistralOllamaLM Studio
How it works

From “where is that data?” to done, in three steps.

  1. 1

    Connect

    Paste a connection URL, pick a database Kiyi found on your computer, or open the sample store. Hosted databases like Supabase, Neon and RDS get the right SSL settings on their own; private ones are reached through SSH, AWS SSM or Kubernetes.

  2. 2

    Look around

    The Overview lists every table. Open one to browse it like a sheet, filter it, ask AI for the rows you mean, group it into a summary or draw it as a chart.

  3. 3

    Change it safely

    Edit cells, insert rows, clean up messy values or redesign a table. Kiyi shows what will happen in plain words, keeps an Undo next to every save, and asks you to confirm before anything changes on production.

Ask AI

Ask for rows the way you'd ask a colleague.

pro customers in germany who spent over 1k

Kiyi turns your sentence into filters you can see, tweak and remove one by one. Nothing runs that you can't inspect.

  • Any provider, or none. Claude, GPT, Gemini, open models on OpenRouter or Groq, or a model running on your own computer with Ollama.
  • Your rows never leave. Only the table's structure is sent, column names and types.
  • Checked before it runs. Anything the AI writes must parse as a single read-only condition.
Ask AI turned a request into three filters and a sort, shown as removable chips
Questions & charts

Ask a question, get a chart.

Type “monthly revenue over the last year” in any language. Kiyi writes one read-only query, runs it, and draws the answer. The SQL is right there to read, change or save.

  • Any result as a chart. Bars, a line, a pie or a single figure, picked from the columns and yours to change.
  • Summaries without SQL. Group a table by any column, dates by day, month or year, and see counts, totals and averages.
  • AI in the SQL editor. Write a query from a description, explain one in plain words, or fix the last error.
A question turned into a SQL query, with monthly revenue drawn as a line chart
Move data

Bring another system's data into yours.

Switching from another app, merging a company you took over, or leaving an old schema behind? Point Kiyi at their database and yours. It works out which table feeds which, how each column fills, and how values change on the way, even when nothing is named alike.

  • A plan you can read. Columns matched by name, synonyms and links; full names split into first and last; codes like “A” turned into “active”. Every step is listed, and AI can refine it.
  • Links that survive. Rows can get new IDs after yours, and every order still points at its customer, even one you already had.
  • All of it, or none of it. Check real rows first, do a test run that rolls back, then move everything in one transaction.
  • Any direction. PostgreSQL, MySQL, MariaDB, SQL Server and SQLite, with time zones, yes/no values and JSON converted on the way.
Moving a legacy CRM's clients into a shop's customers: full names split, codes mapped to statuses, and a preview of the rows as they'll be stored
Data quality New

Find messy data before your reports do.

Imports, forms and years of hand edits leave marks: “Istanbul”, “istanbul ” and “İstanbul” counted as three cities, the same customer entered twice, a price of −5. Open a table's Quality tab, or check the whole database from the Overview, and Kiyi lists what looks wrong, how many rows it touches, and a fix.

  • The usual suspects. Duplicate and copied rows, one value spelled several ways, stray spaces, empty text, invalid emails, outliers, negative amounts, dates in the future or a century ago, columns nobody fills.
  • Fixes you review. “Use ‘Istanbul’ everywhere” or “Trim the spaces” shows its SQL and how many rows it changes first, and runs in one transaction. Nothing changes on its own.
  • See the rows. Show the rows opens the table filtered to the ones behind a finding, so you can fix them by hand when a rule isn't enough.
  • Quick on big tables. Up to 50,000 rows per table are checked, read-only, on your computer. Nothing is sent anywhere.
The Quality tab of a customers table: a city spelled three ways with a one-click fix, and values with extra spaces
Dashboards New

The numbers you check every morning, on one page.

Any chart you make in the SQL editor or from a question can go on a dashboard with Add to dashboard. Keep revenue, orders by status and new sign-ups side by side, and open them all with one click instead of re-running five queries.

  • Fresh when you look. Charts re-run when the dashboard opens, on Refresh, or every few minutes while it's on screen.
  • Arrange it your way. Make a chart full width, move it left or right, open its SQL, or rename the dashboard.
  • Read-only by design. Dashboard queries can only read, so a refresh never changes data, even on production.
  • Share the setup. Dashboards travel with exported connections, so a teammate gets the same charts against their own access.
A sales dashboard with revenue this year, orders by status as a pie, revenue by month as a line and customers by plan
Speed

Fast enough to forget about.

A Rust core and the system's own web view instead of a bundled browser. Measured on a 2021 MacBook Pro with a million-row table.

0.31 sfrom click to window
79 MBmemory at rest
12 MBdownload
~1 msto open, sort or filter a 1M-row table
Everything in one window

Built for people who aren't DBAs.
Loved by the ones who are.

Edit like a spreadsheet

Change a cell and it's saved, with an Undo right there. Production connections wait for an explicit Save. Deleting always asks first.

Finds your databases

Kiyi spots PostgreSQL, MySQL and MariaDB running on your computer or in Docker, and connects in one click.

Structure, in plain words

“Money / exact number”, “Required”, “Unique”, “Links to customers”. Kiyi writes the SQL.

Private networks

SSH tunnels through a bastion, AWS Systems Manager, Kubernetes port-forward or the Cloud SQL Auth Proxy.

Import & export

Any view or query result to Excel, CSV or JSON. Excel and CSV in, all or nothing. Take your connections to another computer, without their passwords.

Safe by default

Production connections are marked in red and read-only unless you say otherwise. Every destructive change is explained in plain language before it runs, and you type the table's name to confirm on production.

A real SQL editor

Schema-aware autocomplete, formatting, readable query plans, history and saved queries with :parameters that ask for their values, results of a one-table query you can edit and save, and the SQL behind every action in Developer mode.

Charts & dashboards

Any result as bars, a line, a pie or a single figure. Group a table by any column without writing SQL, and keep charts side by side on a dashboard.

See the whole database

A diagram of every table and link; functions, triggers, sequences and users with the SQL behind them; and a comparison of two databases with the SQL that would make them match.

Back up & restore

With pg_dump or mysqldump when you have them, Kiyi's own backup when you don't. Restores ask for the database's name first.

Work faster

⌘K to jump to any table, query or connection. Saved views, hidden and frozen columns, totals of the selected cells, copy rows as CSV, Markdown, JSON or INSERTs, and find and replace across a column.

Finds messy data

Duplicates, “Istanbul” spelled three ways, stray spaces, invalid emails, outliers and impossible dates, with a fix you review before it runs.

Speaks your language

Ask in English, Turkish, German or anything else. Excel and CSV files with Turkish and other encodings come in right.

Who it's for

One app for everyone who touches the data.

Operations & support

Look up a customer, fix an address, cancel an order, refund a payment. Search the way you'd describe it and edit like a sheet, with Undo.

Analysts & founders

Ask “revenue by month this year” and get a chart. Keep the ones that matter on a dashboard and export anything to Excel.

Developers

A fast SQL editor with autocomplete, parameters, plans and editable results; schema changes without hand-written DDL; one app for five engines.

Data & migration work

Clean up duplicates and spellings, compare two databases, and move another system's data into yours with links intact.

Structure

Design tables without learning DDL.

Pick a type from a list that makes sense, mark a column required or unique, link it to another table and decide what happens when that row is deleted. Kiyi turns it into the right ALTER statements for your database, in the right order.

  • Rename, add and remove columns, indexes and relationships.
  • MySQL quirks handled, like keeping comments and ON UPDATE when a column changes.
  • Saving a table you didn't change does nothing at all. Tested against real servers.
Editing a column: type, default value, rules and links to other tables
Connect anywhere

From localhost to a private subnet.

Paste a connection URL, pick a SQLite file, or go through a bastion. Every step of the connection test tells you exactly what worked, and what didn't.

  • SSH: password, key file or ssh-agent, with the server's identity remembered and checked.
  • AWS SSM: reach RDS through an EC2 instance with no open SSH port.
  • Kubernetes and Cloud SQL: port-forward to a service or pod, or connect through the Cloud SQL Auth Proxy.
  • Hosted databases: Supabase, Neon, PlanetScale, Amazon RDS, Azure SQL, DigitalOcean and more are recognised, with SSL set up and a pointer to where their details are.
  • Keychain: passwords and keys live in your system keychain, never in a file.
A connection test through an SSH tunnel, every step passing
SQL editor New

Write the query, then fix what it finds.

When a query reads one table and shows its key, its result is as editable as the table itself. Narrow down the rows with SQL, change the cells right there, and Save. No copying ids into an UPDATE by hand.

  • Only what's safe. Joins, totals, DISTINCT and computed columns stay read-only, because they don't map to one row.
  • Saved by key. Changed cells are highlighted until you Save; each row is updated by its primary key in one transaction.
  • Reviewed on production. The exact UPDATE statements are shown before anything is saved.
  • A real editor around it. Autocomplete that knows your schema, :parameters, formatting, readable plans, history and saved queries.
A SQL query's result with two edited cells highlighted and Save and Discard in the status bar

Your data stays yours.

No account.Download, open, connect. That's it.
No telemetry.Kiyi talks to your databases, the AI provider you pick, and GitHub to check for updates. Nothing else.
Open source.MIT licensed. Read every line on GitHub, build it yourself.
Native and fast.A Rust core that streams results straight to the grid.
Download

Get Kiyi v0.6.0

Free. Updates itself in the background and asks before restarting.

macOS

macOS 11 or later

  • Signed and notarized by Apple
  • M1 and later, or Intel

Windows

Windows 10 or 11, 64-bit

  • Installs for you alone, no admin
  • Passwords in Credential Manager

Linux

64-bit, with a desktop keyring

  • AppImage: any distribution, updates itself
  • .deb: Debian, Ubuntu and Mint

The macOS app is signed and notarized by Apple. The Windows installer isn't code-signed yet: if SmartScreen asks, choose More info → Run anyway. On Linux, Kiyi keeps passwords in GNOME Keyring or KWallet; the AppImage updates itself.

Questions

Do I need to know SQL?

No. Browsing, searching, filtering, editing, importing, charting and designing tables all work with buttons and plain words, and you can ask questions in any language. If you do know SQL, the SQL editor is one click away, and Developer mode shows the statements behind every action.

Does the AI see my data?

Kiyi sends the table's structure (column names, types and allowed values) and your request, not your rows. The one exception is planning a data move, where you can choose to show the AI three example rows per table; it's off unless you tick it. You choose the provider, and you can run a model entirely on your own computer with Ollama or LM Studio.

Is it safe to use on production?

Mark a connection as Production and Kiyi makes it read-only by default, outlines it in red, waits for an explicit Save on edits, and asks you to type the table's name before anything destructive. Grid edits run in a single transaction and are rolled back if any row changed underneath you.

How do I reach a database in a private network?

Under “Connect through”, choose SSH to go through a bastion host, AWS SSM to go through an EC2 instance using Systems Manager, Kubernetes to port-forward to a service, or Cloud SQL for the Cloud SQL Auth Proxy. The connection test shows each step: tunnel, sign-in, forwarding, database.

Which databases are supported?

PostgreSQL, MySQL, MariaDB, SQLite and SQL Server (2017 and later, and Azure SQL). Kiyi's core is built so new engines plug in without changing the app.

Can Kiyi move data from another system into mine?

Yes, between PostgreSQL, MySQL, MariaDB, SQL Server and SQLite in any direction, even when the tables look nothing alike. Choose Move data on the Overview, pick the other database, and Kiyi suggests which table and column goes where. You adjust the plan, check real rows, try a test run that keeps nothing, then move it all in one transaction. Plans can be saved and run again later.

How does the data quality check work? Will it change my data?

It reads up to 50,000 rows of a table (10,000 per table when checking the whole database) and looks for duplicates, values spelled several ways, stray spaces, invalid emails, outliers, negative amounts and impossible dates. It only reads. Fixes are offered as buttons that show the SQL and the number of rows first, and nothing runs until you confirm.

Can I edit the result of a SQL query?

Yes, when the query reads a single table and includes its primary key, like SELECT id, name, email FROM customers WHERE …. Edited cells are highlighted until you press Save, which updates each row by its key. Results of joins, groupings and computed columns stay read-only.

Do dashboards update by themselves?

They re-run their queries when you open them and when you press Refresh, and you can have a dashboard refresh every few minutes while it's on screen. Dashboard queries are checked to be read-only, so refreshing never changes data.

Can I try it without a database?

Yes. Choose “Try the sample database” on the welcome screen and Kiyi opens a small store with customers, products and orders to browse, chart and edit. It lives on your computer and you can recreate it any time.

What language is Kiyi in?

The interface is in plain, short English. You can ask questions and give AI requests in any language and get answers back in it, and Excel and CSV files in other languages' encodings open correctly.

What does it cost?

Nothing. Kiyi is free and open source under the MIT license. If you use AI, you pay your provider directly, or nothing at all with a local model.

Make your database feel like home.