DataLine
DataLine is an open-source AI tool that turns plain-English questions into SQL, tables and charts across Postgres, MySQL, Snowflake, CSV and Excel. It runs entirely on your own machine, so your data never reaches a vendor server.
What is DataLine?
DataLine is an AI data analysis and visualisation tool built around a single idea: you should be able to ask a question of your data in plain English and get a table, a chart or a dashboard back, without writing a query and without your data leaving your machine.
Its homepage names the audience in one line — non-technical people who need to query data, and developers looking for a best-in-class text2sql solution. The tool sits between those two groups. For the first, it removes the wait for a data team and the hunt through table schemas and column names. For the second, it drafts and executes SQL fast enough to be genuinely useful when exploring an unfamiliar database.
The architecture is the differentiator. DataLine is self-hosted: you install it as a binary on Windows, macOS or Linux, through Homebrew, or as an official Docker image, and it serves a web interface on http://localhost:7377. Conversations, connections and results are stored on your own machine in SQLite. The vendor says it owns no server for this, and that even the marketing site is statically hosted. By default the language model never sees your rows: it receives only descriptions of your data — table names, column names and their types. You can lift that restriction when the data is not sensitive, which lets the model read sample rows and results.
It connects to Postgres, MySQL, SQLite, Microsoft SQL Server, Azure SQL Server, Snowflake and BigQuery, and reads flat files including CSV, Excel workbooks — each sheet ingested as a separate table — and SAS sas7bdat. Beyond generating SQL it charts results in natural language, lets you edit and re-run a generated query, builds dashboards and triggers, and keeps a knowledge base of trainable examples.
DataLine is free and open source under the GPL-3.0 licence, with around 1,600 stars on its public repository. You supply your own OpenAI API key, and an alternative base URL can be configured; local model support is announced as coming soon. It is published by DataLine B.V., a two-person team — Rami Awar and Anthony Malkoun — who prototyped in April 2023 and open-sourced the project in February 2024.
What it does
- Turn a plain-English question into an executable SQL query
- Generate charts and dashboards from the same conversation
- Connect to Postgres, MySQL, SQLite, Microsoft SQL Server, Azure SQL Server, Snowflake and BigQuery
- Query flat files: CSV, Excel workbooks, SQLite databases and SAS sas7bdat
- Edit the generated SQL, save it and re-run it later
- Build dashboards and triggers on saved results
- Keep a knowledge base of trainable examples to sharpen later answers
When to use DataLine / When not to
A quick filter to help you decide if DataLine is the right fit.
When to use DataLine
- Non-technical staff who need answers from a database but do not write SQL
- Backend developers drafting queries or exploring an unfamiliar schema
- Data analysts wanting a faster first draft without giving up control of the query
- Businesses handling sensitive data that cannot be sent to a third-party cloud
- Researchers and statisticians working with CSV, Excel or SAS sas7bdat files
When not to use DataLine
- Teams wanting a ready-to-use hosted service with nothing to install
- Organisations needing multi-user accounts and role management
- Buyers who require a signed DPA or SOC 2 / ISO certification
- Anyone looking for a mobile app or a public API to build on
- Users unwilling to send even table and column names to a third-party model
How to use DataLine
A typical end-to-end flow, from setup to results.
- Pick an install route: a downloadable binary, Homebrew, or the official Docker image
- On Windows, unzip the release archive and run DataLine.exe, clicking past the 'Windows protected your PC' warning
- On macOS or Linux, run brew tap ramiawar/dataline, then brew install dataline, then launch dataline
- With Docker, run the published container command, mapping port 7377 and the dataline volume
- Open http://localhost:7377 in your browser
- Paste your own OpenAI API key into the settings, optionally pointing the base URL at another endpoint
- Create a connection to a database, or import a CSV, Excel, SQLite or sas7bdat file
- Start a conversation and ask your question in plain English
- Inspect the generated SQL, edit it if needed, then save it and re-run it
- Ask for a chart, refine the chart query, and assemble results into a dashboard
Pros & Cons
Pros
- Free and open source under GPL-3.0, so the code can be audited
- Data never reaches the vendor: queries run and results are stored on your own machine
- The language model is blind to your rows by default, seeing only table and column names
- Broad source coverage: seven databases plus CSV, Excel and SAS files
- Installs in one command with Homebrew or Docker, or as a double-click binary
- Serves non-technical staff and developers from the same interface
- No subscription: the only running cost is your own model usage
Cons
- No hosted version: you install and maintain it yourself
- No terms and conditions published anywhere on the site
- You must bring, and pay for, your own OpenAI API key
- Authentication is single-user and unavailable when running the standalone executable
- No DPA, no certification, and no explicit GDPR statement despite a Dutch company name
- Local model support is still announced only as coming soon
- The repository carries a call for maintainers and the site build dates from May 2025
Pricing & Plans
DataLine is free of charge. The software is distributed under the GPL-3.0 open-source licence, there is no pricing page and no paid tier of any kind, so no entry price and no currency can be quoted. The only cost attaches to the language model: you supply your own OpenAI API key and usage is billed to you directly by that provider, not by DataLine. No free trial is advertised, for the straightforward reason that there is nothing to trial — the complete product is free.
Data, GDPR & hosting
A consolidated view of how DataLine handles your data.
GDPR overview
There is no mention of the GDPR anywhere on the site. Neither the privacy policy nor any other page names the regulation, claims compliance, or denies it. The wording is nonetheless GDPR-adjacent: the policy of 21 June 2024 designates DataLine as data controller and grants rights of access, correction and deletion, exercisable through the application settings or by writing to support. The company name in the footer, DataLine B.V., is a Dutch legal form and so points to an EU establishment. What is missing is everything formal — no Article 27 representative, no named data protection officer, no data processing agreement, and no subprocessor register. Sentry and LangSmith are named in passing in the sharing section, which is disclosure but not a maintained list.
Who owns the data?
DataLine publishes no terms and conditions, so no contract assigns ownership of the content you handle. The only governing document is the privacy policy of 21 June 2024, which names DataLine as data controller for the limited personal data it processes: contact details you volunteer when asking for support, and error reports sent through Sentry if you enable them. Your business data falls outside that scope because it never reaches the vendor. Queries run on your own device and results are stored locally in SQLite. The policy states plainly that the company does not have access to or store this data, and that nothing is kept on its servers.
Reuse rights
Because no terms and conditions exist, nothing restricts what you do with the data you query or the outputs you generate: both stay on your own machine in a local SQLite store, and the vendor claims neither access nor ownership. Results are explicitly meant to be copied out to build reports quickly. The software itself is a separate matter: it is distributed under the GPL-3.0 licence, so reusing, modifying or redistributing the code carries that licence's obligations. Only two third parties can ever see fragments, and both are under your control — Sentry receives error reports after you consent, and LangSmith receives tracing data only if you configure it.
Data retention & training
Hosting summary
There is no vendor hosting to describe: DataLine processes and stores everything on the machine where it runs. Conversations, connections and results live in a local SQLite database, and in Docker deployments they persist in a named volume across upgrades. The FAQ puts it bluntly — no servers, no cloud, no risk of leaks, the company does not even own a server, and its website is statically hosted. The practical consequence is that the jurisdiction governing your data is simply that of your own hardware; the vendor declares none, which is why no hosting country or region is recorded here. Two third parties may receive fragments, both optional and both under your control: Sentry, for error reports, only if you enable it and consent, and LangSmith, for tracing, only if you configure it. One distinction is worth keeping straight: the dataline.app website itself resolves to a United States address on Render's network, but that concerns the pages you are reading, not the data you query.
Things to keep in mind
Risks and trade-offs to weigh before adopting DataLine.
- Switching off the data-hiding setting sends real rows to a third-party model: a one-click decision with lasting consequences
- Even in the default mode your schema travels to OpenAI, and table and column names alone can be revealing
- AI-generated SQL can be confidently wrong, and a plausible number is far harder to catch than an error message
- Leaning on generated queries can quietly erode your own SQL fluency over time
- With no terms and conditions, there is no service commitment and no stated liability to fall back on
- Everything lives on your machine, so backups, disk encryption and a lost laptop become entirely your problem
- The standalone executable has no authentication: anyone with access to the computer reaches your connections
Setup & Integrations
Technical difficulty
Low for local use. Three routes are offered: a double-click binary for Windows, macOS or Linux, a two-command Homebrew install, or a single Docker command. The interface then opens at http://localhost:7377. The one real prerequisite is an OpenAI API key, which you obtain yourself and paste into the settings; Windows users must also click past a security warning. Difficulty rises only if you expose the instance on a domain, which needs ALLOWED_ORIGINS set for CORS and AUTH_USERNAME and AUTH_PASSWORD for basic authentication. Running from source requires a Python FastAPI backend and a React frontend.
Deployment
Integrations
Supported languages
Behind DataLine
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement DataLine.
Frequently asked questions
Is DataLine free?
Where is my data stored?
Does the language model see my data?
Which data sources does it support?
How do I install it?
Can I secure a self-hosted instance?
Does DataLine train on my data?
Is there a mobile app or a public API?
Who is behind DataLine?
Should you pick DataLine?
DataLine answers a narrow question well: how do you let anyone interrogate a database in plain English without handing that database to someone else? Its answer is to invert the usual arrangement. Nothing is hosted, nothing is uploaded, and the language model is shown your schema rather than your rows unless you decide otherwise. For a team sitting on data it cannot export — or simply unwilling to — that is a materially different proposition from a cloud assistant, and the seven database connectors plus CSV, Excel and SAS support make it practical rather than theoretical.
The cost of that design is that you carry it. You install the software, you obtain and pay for an OpenAI API key, you secure the instance, and you back up the SQLite file that holds your work. Authentication is basic and single-user, and it disappears entirely if you run the standalone executable. The legal surface is thin too: a privacy policy from June 2024, no terms and conditions at all, no processing agreement, and no mention of the GDPR despite a Dutch company name — a gap that will matter to procurement long before it matters to an individual user.
Momentum is the other open question. The project has roughly 1,600 stars and a handful of contributors, but its README asks for maintainers and the website build has not moved since May 2025. Local model support, which would close the last gap in the privacy story, remains announced rather than shipped. Read as a free, auditable tool for people who want their data to stay put, DataLine is a genuinely useful thing to install; read as a supported product to standardise a company on, it asks for more trust than its current documentation extends.
- Choosing a selection results in a full page refresh.
- Opens in a new window.