SQLore
FeaturesPricingDownloadAboutBlogContact
Start freeStart for free
SQLore
HomeFeaturesSolutionsPricingDownloadCompareFAQAboutBlogContact

SQLore © 2026

TermsPrivacyRefunds
Download for macOSDownload for WindowsSign in
  1. Home
  2. Solutions
  3. The category, explained
The category, explained

The AI Database Assistant That Runs On Your Machine

An assistant that reads your schema, writes the query, runs it read-only and shows you exactly what it did — with no rows leaving your network.

Download SQLoreSee all features

What is an AI database assistant?

An AI database assistant is a tool that turns a plain-English question into SQL, runs it against your database, and gives you back the answer — explained in words, drawn as a chart or a table, and accompanied by the query it used. SQLore is an offline-first AI database assistant for PostgreSQL and MySQL: it introspects your schema, generates read-only SQL, and runs the whole loop on a local model unless you opt into a live one.

What separates an assistant from an autocomplete

Editor autocomplete finishes the line you are already writing. An assistant answers a question you have not worked out how to ask in SQL yet — which means it has to do three things autocomplete never does: find the right tables in a schema it was not trained on, run the query, and read the result well enough to notice when it is wrong.

  • It resolves the question against your actual schema, not a generic model of one.
  • It executes, so an answer is rows rather than a query you still have to verify.
  • It reads the result back to you and draws it — a figure, a bar chart, a donut, a table — rather than leaving you to interpret a grid.
  • It retries: a bad join shows up in the result, and it fixes it before you see anything.
  • It shows the SQL, so the answer is checkable rather than trusted.

Why the schema is the hard part

Any current model can write syntactically valid SQL. Almost none can write correct SQL against a two-thousand-table warehouse it has never seen, because correctness is a question about your schema and not about SQL. SQLore introspects the database once and caches what it finds — tables, columns, keys, distinct values and statistics — then searches that index locally before a single token of SQL is generated. That is why a filter lands on the column that actually holds the value.

What an assistant should never be allowed to do

Write. There is no question worth answering that requires an UPDATE, and an assistant with write access is one hallucinated statement away from an incident. SQLore parses generated SQL to an abstract syntax tree and validates it down to a single read-only SELECT before it goes near your database, then executes it on a read-only connection with a statement timeout. That is a parser, not a blocklist of dangerous words — the distinction matters, because a blocklist is a list of the strings someone thought of.

Offline is a posture, not a feature toggle

Most assistants in this category are a web app with a form for your connection string, which means your credentials and your schema live on someone else's server. SQLore is a desktop application: the credentials stay on your machine, the query runs from your machine, and with a local model configured, nothing leaves your network at all. Live cloud models are something you switch on for a hard question, not something you have to find and switch off.

Frequently asked questions

What is the difference between an AI database assistant and a SQL generator?

A SQL generator returns a query and stops. An assistant runs the query against your database, reads the result, corrects itself when the result shows the query was wrong, and hands back an answer you can actually use — described in plain English and charted, with the SQL there to check. The practical difference is who does the work after the query is written: with a generator it is you, every time.

Do I need to be a SQL developer to use one?

No, but you need to be able to recognise a wrong answer. SQLore shows the query, the row count and the execution time with every result precisely so that a non-specialist has something to sanity-check against, and a specialist can read the SQL directly.

Is an AI database assistant safe to point at a production database?

It is safe if it cannot write and cannot run long. SQLore validates every generated statement to a single read-only SELECT and executes it on a read-only connection with a statement timeout. Pointing it at a read-only replica with a read-only role is still the configuration we recommend, because two independent guarantees are better than one.

More answers on the main FAQ, or ask us directly.

Related

For people who already write SQL

An AI SQL Assistant That Shows Its Working

Read more

The technique

Natural Language To SQL, Done Against A Real Schema

Read more

Local models

A Local AI Database Assistant, Model Included

Read more

Every page in this section covers a different question about the same product. See all solutions.

SQLore platform

Stop Writing The
Same Three Joins

Start free, upgrade when you are asking enough questions for it to matter.

Start for freeContact sales
sqlore.com

New chat

Search chats

TODAY

Orders by month

PREVIOUS 7 DAYS

Top products by revenue

PREVIOUS 30 DAYS

Repeat customer rate

Revenue by category

Average order value

Refunds by reason

Customers by city

Cart abandonment

Slowest shipping lanes

Discount impact on margin

Inventory turnover

Order statistics

SQLoredesktopretail_db · default⌘K

Order statistics for this year

Reasoning — 2 steps, 2 queries

Total orders · Delivered · In transit · Cancelled

1 row · 10 ms

Stat

15,600

Total orders

13,942

Delivered

1,412

In transit

246

Cancelled

orders by category

4 rows · 11 ms

LineDonutBar
8,4006,3004,2002,1000
ACCESSORIESPERIPHERALSFURNITUREDISPLAYS

There are 15,600 orders in retail_db for the year to date.

  • Status: 13,942 have been delivered and 1,412 are still in transit.
  • Category: most sit in Accessories (8,420), followed by Peripherals (4,180), Furniture (2,140) and Displays (860).

Revenue by payment method

Reasoning — 1 step, 1 query

Ask about your database…

Answers come from SQL run against your database.