Saytabase docs: the voice database, explained.

Getting started

Create your first database in under a minute — no setup, no spreadsheets.

1. Go to the homepage and tap the microphone, or type a sentence describing something you want to keep track of. Example: “Met Dana at the coffee shop, she wants a 3-bedroom near downtown, budget around 600k.”
2. Saytabase shows you a preview: the columns it picked (name, need, area, budget) and the row it filled in.
3. Tap Save. If you're not signed in yet, create a free account — your preview is kept and saved to your account automatically.
4. Keep adding. Every new note you say or type becomes another row in the same database.

You don't need to plan columns ahead of time. Just describe things the way you normally would.

Adding information by voice or text

Talk naturally. Saytabase sorts each detail into the right column.

Open any saved database and use the speak box at the top. Hold to talk, or type. Saytabase reads what you said, matches each detail to an existing column, and adds the row.

• Mention something new (for example a phone number you've never recorded before) and a new column is added automatically. Older rows simply leave it blank.
• Saying “price” when your column is called “msrp” still works — each column has a plain-English description and alternate names it recognizes.
• Dates, numbers, and yes/no answers are stored as the right type, so sorting and filtering behave as you'd expect.

Working with your rows

View, filter, sort, edit, and export — all from the browser.

• Column pills at the top filter rows by whether a value is present. A separate sort menu orders rows, and date columns support date ranges.
• Tap a row to open it. Edit any value, fill in an empty column with “Add detail”, add a brand new column, or type an update in your own words and let Saytabase apply it.
• Make a copy of a row, copy its summary, or delete it (you'll be asked to confirm).
• Tap the pencil on a column to give it a description and alternate names. The sparkle button writes these for every column automatically by looking at your data — check any you've hand-edited afterward.
• Download everything as CSV whenever you like.

Asking questions about your data

Ask in plain words and get answers grounded in your actual rows.

Use Ask It Back to ask a question like “Which clients want a home office?” or “What's still left to do before Saturday?”. The answer is based only on rows you've saved, and matching rows are highlighted. If the answer isn't in your data, it says so instead of guessing.

To ask from inside ChatGPT, Claude or another assistant you already use, connect it once using the MCP connection below.

Sharing a form

Collect entries from customers, guests, or teammates with a simple link.

1. Open a database and find “A form you can share”.
2. Pick which questions to show, rename them, add placeholder hints, mark them required, and set their order. Only the questions you pick are shown.
3. Turn the form on and copy the link. Anyone with it can submit; each submission becomes a new row.

Form links can only add rows. People who fill in your form can never see existing entries. You can turn the form off at any time.

Connecting ChatGPT, Claude and other AI assistants (MCP)

Let your assistant read and add to your databases after you approve it.

Saytabase hosts a Model Context Protocol (MCP) server. Add this address as a custom connector in a compatible assistant and sign in with your Saytabase account when asked:

https://saytabase.xyz/mcp

1. ChatGPT: open connector settings, create a custom connector named Saytabase, paste the address above, publish it, sign in, then turn it on in your chat.
2. Claude: open Settings → Connectors, add a custom connector with the address above, complete sign-in, and enable it in your conversation.
3. Cursor and other MCP clients: add a remote server using the configuration below and complete sign-in when prompted.

{
  "mcpServers": {
    "saytabase": {
      "url": "https://saytabase.xyz/mcp"
    }
  }
}

• list_vaults — see the databases on your account.
• get_vault_schema — see a database's columns, with descriptions and alternate names.
• query_vault — keyword search across the 200 newest rows; returns up to 50 results (10 by default).
• store_memory — save a new row from flat text, number or yes/no values. Unknown names create new columns.

These tools don't edit or delete existing rows. Custom connector availability depends on your assistant plan. Full walkthrough: https://saytabase.xyz/mcp-server

REST API for developers

Use any database as a live backend for a website, script, or automation.

Each database has its own address and secret key, shown on the database page under the developer section. Send the key as a Bearer token (recommended), an x-saytabase-key header, or a ?key= parameter. Keep the key on a server — never in public web pages.

https://saytabase.xyz/api/public/v1/<database-id>/records
https://saytabase.xyz/api/public/v1/<database-id>/schema

• GET /records — read rows. Options: limit (default 100, max 500), q (keyword search), format=csv.
• POST /records — add a row. Send flat JSON fields, or {"text": "..."} to let Saytabase structure a sentence for you. Optional: summary, source.
• DELETE /records?id=<row id> — delete one row.
• GET /schema — list columns with type, description, and alternate names.

# Read rows
curl "https://saytabase.xyz/api/public/v1/<database-id>/records?limit=20" \
  -H "Authorization: Bearer <secret-key>"

# Add a row
curl -X POST "https://saytabase.xyz/api/public/v1/<database-id>/records" \
  -H "Authorization: Bearer <secret-key>" \
  -H "Content-Type: application/json" \
  -d '{"name":"Ada Lovelace","email":"ada@example.com","status":"new"}'

const res = await fetch("https://saytabase.xyz/api/public/v1/<database-id>/records", {
  method: "POST",
  headers: { Authorization: "Bearer <secret-key>", "Content-Type": "application/json" },
  body: JSON.stringify({ text: "Booked Sam for a brake check Tuesday at 9" }),
});
const { ok } = await res.json();

Privacy and security

Your data stays in your account, with separate keys for separate jobs.

• Each account's databases are kept separate; other users can't see them.
• Three different kinds of access, each limited to its job: assistant connections (sign-in approval), developer secret keys (one database), and form links (add-only).
• When you ask a connected assistant a question, the relevant rows are sent to that assistant and its own terms apply.

Read the full policies: https://saytabase.xyz/privacy and https://saytabase.xyz/terms.

Frequently asked questions

Quick answers to common questions.

Do I need to know databases or code?

No. You talk or type like normal; Saytabase creates and organizes the columns. Developer features are optional.

What happens if I mention something new later?

A new column is added. Existing rows keep working and just show that column as empty until you fill it in.

Can I use it on my phone?

Yes. It works in any modern phone or desktop browser, including voice capture.

Which AI assistants work with it?

Any assistant that supports remote MCP connectors, including ChatGPT, Claude and Cursor. You can also ask questions directly inside Saytabase.

Can I get my data out?

Yes. Download any database as CSV, or read it through the REST API.

How do I get help?

Use the contact form at https://saytabase.xyz/contact.