Only the schema is read
dblm learns tables, columns, keys and relationships; your rows stay where they are.
New · Pin any answer to a live Home board
You're one ask away from the insight you're looking for
Ask a question the way you'd ask a colleague. dblm writes the query, runs it on your own databases, and hands you the rows, along with the exact query and a sentence on what it found.
you asked
✓ Answered from bank
Which 10 branches took in the most deposits in 2025, and how much did each bring in?
- Read the schema
- Wrote the query
- Ran it · 10 rows
- Drew the chart
- Summarised
Top 10 Branches by Deposits
total_deposits by branch_name · USD · chosen by dblm from the shape of the result
$10M
$77.92Mtop-10 total
- Seattle Downtown17.8%
- Portland Pearl13.6%
- Denver LoDo12.2%
- Chicago Business11.6%
- Austin Congress10.7%
- Dallas Business7.9%
- Phoenix Central7.1%
- San Diego Bay6.7%
- Boise Main St6.3%
- Sacramento East6.1%
| branch_id | branch_name | total_deposits |
|---|---|---|
| 7 | Seattle Downtown | 13,874,215.33 |
| 3 | Portland Pearl | 10,612,508.35 |
| 11 | Denver LoDo | 9,468,331.53 |
| 1 | Chicago Business | 9,027,963.97 |
| 9 | Austin Congress | 8,351,228.00 |
| 5 | Dallas Business | 6,188,745.39 |
| 12 | Phoenix Central | 5,536,902.63 |
| 2 | San Diego Bay | 5,241,389.07 |
| 14 | Boise Main St | 4,907,516.85 |
| 6 | Sacramento East | 4,715,044.34 |
SELECT b.branch_id, b.name AS branch_name,
ROUND(SUM(t.amount), 2) AS total_deposits
FROM branches b
JOIN accounts a ON b.branch_id = a.branch_id
JOIN transactions t ON a.account_id = t.account_id
WHERE t.txn_type = 'deposit'
AND t.posted_on >= '2025-01-01'
AND t.posted_on < '2026-01-01'
GROUP BY b.branch_id, b.name
ORDER BY total_deposits DESC
LIMIT 10
dblm says
Seattle Downtown leads at $13.87M, $3.26M ahead of Portland Pearl. They are the only two branches past $10M.
Key observations
- +$3.26M vs #2
Top performers. Seattle Downtown leads by a wide margin, nearly $3.26M more than Portland Pearl. They're the only two branches above $10M. - $13.87M → $4.72M
Distribution. Deposits across the top ten run from $13.87M at Seattle Downtown down to $4.72M at Sacramento East.
Hover or tap an observation to see it on the chart.
❯
Ask a follow-up…
One engine for every question. Close to your data, never inside it.
12.8s
From question to answer
The exchange above: three tables joined, the query written, the rows returned, charted and summarised.
4×
Ways to read every answer
A bar chart, a pie, the raw rows, and the query behind them, shown verbatim so any number can be checked.
0
Rows sent to the model
dblm reads the schema: tables, columns, keys, relationships. Your rows stay exactly where they are.
Speaks to the databases you already run
Point it at a connection, and the schema is indexed and annotated. One question can span several sources at once.
PostgreSQL
MySQL
SQLite
MongoDB
ClickHouse
OpenAPI as tables
Swagger as tables
REST APIs as tables
Postgres here, MongoDB there, a REST API in between, all in one question.
Why askDBLM
A quick question shouldn't take three days
The answer was already in your database. It just had to become a ticket first: filed, queued behind eleven others, answered by whoever can write the query, and returned as a CSV that answers yesterday's version of the question.
Waiting on the data queue
Every follow-up is a new ticket, and every ticket waits its turn.
STATUS: QUEUED · 11 AHEADDATA-427P33 days
Quick question: churn by cohort?
filed Tue 09:14 · queued behind 11 others
DATA-431P32 days
Same thing, but last month
follow-up · waits its turn again
DATA-439P3open
Actually, split by region too
the question that would have been a reply
churn_v3_FINAL(2).csv (answers yesterday's question)
Asking dblm
Ask, read the answer, ask the follow-up. It remembers what you were just talking about.
STATUS: ANSWEREDONE QUESTION, ANSWERED● OK
question asked09:41:02
query written10 lines
tables joined4
answer returned09:41:14
rows returned10
tickets filed0
ask again any time. there is no queue.
A new interface
Software is learning what people mean
For decades, people learned software: where the menu is, which filter to set, how to write the query. Now the software can do the learning. The interface to data is becoming a sentence.
Menus · then
Click
You learn the app. Every question is a path through forms and menus someone designed in advance.
Dashboards · lately
Filter
You learn the data model. The answer is in there, if you know which filters, fields and charts to combine.
Questions · now
Ask
The app learns you. Say what you want to know in your own words and get the answer itself, not a report to dig through.
Your users shouldn't need to know your schema, or that you have one.
semantic · conversational · grounded in your real data
In your app
Give your users the power to talk to their data
Put dblm inside the product you already ship, whether that's a web app or an iOS or Android app. Your users ask in their own words, and the answer arrives as a chart, a table or a sentence, right where they are.
iOS app
Where did my money go in December?
Your top three in December:
- 1Groceries$287.79 · 3 purchases
- 2Dining$259.78 · 5 purchases
- 3Shopping$170.38 · 1 purchase
Ask about your money…↑
Web app · React
AAcme Bank
AccountsLoansCustomersAsk your dataSettings
How have deposits moved over the last five years?
Deposits dipped 12.6% in 2022 and have climbed every year since. 2025 is 11% above 2021.
$82.7M
$72.3M
$77.5M
$84.4M
$92.0M
20212022202320242025
Ask a follow-up…↑
Android app
Which branches lent the most this year?
The two business centres lead for 2025:
| Chicago Business | $1.40M |
| Dallas Business | $1.08M |
| Seattle Downtown | $678K |
| Phoenix Central | $384K |
Ask about your loans…↑
Your appiOS · Android · WebThe question, in your user's words.
Your backendYour auth, your rulesDecides who may ask what.
dblmUnderstands and answersWrites a read-only query, runs it, charts and summarises.
Your dataDatabases & APIsRead, never changed, and never shipped to the device.
Organizational knowledge
It knows your business the way your team does
Your definitions, your rules and the answers you already trust. Every question is answered with all of it in mind.
Learns your dataSpeaks your termsCorrects its own queries
Dashboards
Design the board your team actually reads
Every answer can be pinned as a report. Drag them onto your board, arrange and resize them, pick the chart that tells the story, and each one keeps itself up to date.
01Pin any answer
Ask in plain English; when the answer is worth keeping, pin it. The query comes with it.
02Arrange it your way
Drag, resize and re-chart each report (bar, line, pie or table) until the board reads right.
03Always current
Each tile re-runs its stored query on its own schedule. No model in the loop, so the numbers are exact.
Security
Read-only. Not "should not". Cannot.
dblm cannot write to your data. Every query it writes, whether SQL, a MongoDB pipeline or a REST call, passes a read-only guard for its own language before it runs. Anything the guard doesn't recognise as a read is refused. There is no flag that turns this off.
The query
written for your engine
The read-only guard
SQLreads only
MongoDBfind · aggregate
RESTGET only
Your data
read, never changed
INSERT · UPDATE · DELETE · DROP · ALTER · TRUNCATE ·
$out · $merge · POST · PUT · PATCH …
writes of every kind are refused before they ever reach your data.
Private columns
Mark one private and it never reaches a prompt, a query or a page.
Your own instance
Every organization gets a dedicated instance of its own, not a shared pool.
Auditable
Every question, query and result is logged per user, so you can retrace any number.
Budgeted
Each deployment has its own spend ceiling; cost cannot run away quietly.
Or self-hosted
Prefer to keep everything in-house? Run dblm entirely inside your own network.
Tailored to your team
The whole machine, in one grid
Everything your team needs to get answers and keep them.
Ask in plain English
Get rows, a chart, and the query it wrote, every time.
It remembers the thread
Say "just last month" or "split by region" and it knows what you were talking about.
One question, many databases
Postgres here, MongoDB there, a REST API in between.
Reusable questions
Save a good question with blanks in it; anyone on the team can run it safely.
Answers that stay fresh
Pin an answer and it refreshes on a schedule. There's no model in the loop, so it stays cheap and exact.
Share it anywhere
The same answer as a dashboard, an image, or a plain CSV for the spreadsheet.
Hide what's sensitive
A hidden column disappears from prompts, results and charts alike.
Every question, searchable
Find how someone got a number six weeks ago.
Viewer, editor, owner
Each with their own audit trail, never a shared login.
Plans
Start free. Run it your way.
Start free on a hosted instance, bring your whole team with Enterprise, or run it inside your own network, also for free.
Free Start here
for getting answers from day one
$0/ month
No card needed
- 100 queries every month
- A dedicated instance of your own
- Your own address at your-org.askdblm.com
Enterprise
for teams that run on their data
$10/ user / month
Everything in Free, for your whole team
- Roles, with a per-user audit trail
- Private columns for sensitive data
- A spend ceiling per deployment
Self-hosted
for data that can't leave your network
$0/ forever
On your own infrastructure
- Runs inside your own network
- Same read-only guarantees
- Nothing leaves your walls
FAQ
Before you ask
Anything else? Sign up and ask your own database. It's free.
Do I need to know a query language?
No, that's the point. You ask in plain English. If you do know one, every answer shows the query it ran (SQL, a MongoDB pipeline or an API call), so you can check the working.
Which databases can it talk to?
PostgreSQL, MySQL, SQLite, MongoDB and ClickHouse today, plus REST APIs described by an OpenAPI or Swagger spec, which surface as queryable tables too. One question can span several of them at once.
Can it change or delete my data?
No. dblm is read-only by construction: every generated query is checked by a read-only guard for its own language (SQL, MongoDB or a REST API) before it runs, and there is no way to run a write through it. See the limits.
Where does my data go?
Your rows never leave your database. dblm reads schema metadata (tables, columns, keys, relationships) and sends only that, plus your question, to the language model. You can mark columns private so they are hidden even from the model.
How long does setup take?
About a minute. Pick an organization name, create your password, and you'll be dropped into a fresh instance. Add a connection and index it, and you're asking questions.
What does the free plan include?
100 queries every month on a dedicated instance of your own, with no card and no time limit. When your whole team is asking, Enterprise is $10 per user per month.
Can I run it on my own server instead?
Yes, and self-hosting is free. dblm can run entirely inside your own network, with the same read-only guarantees, so your data and your questions never leave your infrastructure.
Get started
Your own dblm, in about a minute
Free for up to 100 queries a month, no card. A dedicated instance, and every answer shows its working.
- +0:00
Your instance is created, dedicated to your organization.
- +0:10
You're in the panel at your-org.askdblm.com.
- +0:30
Add a connection; the schema is indexed and annotated.
- +1:00
First question answered, with the query shown.
Already have an instance? Sign in to your portal →
Start free
Free plan · 100 queries a month · ready in about a minute
Your dblm is ready
Sign in at the portal with the password you just chose, and open your instance from there. Its address is . It signs you in with that same account, so there's no second password.