BigQuery,
with everything you wished it had.
Querylab.io is a SQL IDE for BigQuery, in your browser or as a desktop app.
Free. The Playground needs no sign-in.
-- cost shown before you run
SELECT *
SELECT
FROM `ace-analytics.warehouse.events`
WHERE event_date BETWEEN '2019-01-01' AND '2024-12-31'| event_date | user_id | event_name | |
|---|---|---|---|
| 1 | 2024-12-31 | u_8f3c21 | checkout_completed |
| 2 | 2024-12-31 | u_44b7e9 | page_view |
| 3 | 2024-12-30 | u_a01d5c | add_to_cart |
| 4 | 2024-12-30 | u_39cc04 | session_start |
-- new accounts per week and country
SELECT DATE_TRUNC(signup_date, WEEK) AS week,
country,
COUNT(DISTINCT user_id) AS signups
FROM `ace-analytics.warehouse.users`
GROUP BY week- Write
- Completions, go to definition and hover docs that know your schema
- Check
- Errors and warnings before you run, each with a fix
- Run
- On-demand or slots, whichever costs less
- Debug
- Click a step of the query plan to see the SQL behind it
- Limits
- GB caps on every query, including the ones Claude Code runs
Where to start
- Playground
- The full IDE on a sample project. Open it without signing in.
- Your projects
- Sign in with Google and query the projects you already have. BigQuery returns rows straight to your browser. The desktop app keeps your sign-in in the OS keychain.
- Learn
- Learn BigQuery is a free course of 76 lessons. They run in your browser and need no account.
BigQuery charges by the byte.
Querylab.io shows it in dollars as you type.
See the cost while you write
Every query, estimated as you type: dollars, not bytes. The per-CTE breakdown points at the expensive part.
Budgets per user, not per project
BigQuery caps every user with the same quota. Querylab.io sets a limit per person: 10 TiB a day for your senior analyst, 500 GiB for the new hire. Over-budget queries stop before they bill.
Test on a sample
Set a Tablesample percentage and the query reads a sample of each table. BigQuery can't sample views, and a table up to about 168 MiB is one block, so those are read in full. Querylab.io prices the sample and the full query, so you see what you saved.
SELECT * FROM `warehouse.events`80 TiB · ~$500SELECT user_id, ts FROM `warehouse.events` TABLESAMPLE SYSTEM (0.01 PERCENT)8.0 GiB · $0.05Slots or on-demand? Both prices, before you run.
On-demand comes from a dry run. The slot price is predicted from the query's complexity and earlier runs.
Set routing to Auto and each query takes the cheaper of the two on its own.
The Insights panel, LIMIT hints and storage costs are on the All features tab.
An agent that reads your warehouse before it writes your SQL.
You see every tool it calls, and its work lands in this window as a diff, a price, a new tab or a chart. Watch it work:
It looks your schema up first
Prism AI has to search your real tables and columns before it writes a line of SQL. It is not allowed to guess a column name from training data.
A price on the button
Everything it writes gets a BigQuery dry run. In Approve mode the Run button shows the estimated cost and the row cap before anything executes.
Your own AI subscription
On desktop it drives the agent already on your machine (Claude Code, Codex, Gemini CLI, Copilot or Cursor) through your subscription, without routing your SQL through us.
It sees your rows only if you let it
Schema is shared so it can write correct SQL. Result data is not: the default is to ask you each time, and you can turn it off for good.
Or skip the chat: put the cursor in your SQL, hit , and describe the change. The edit is parsed and checked before you're shown the diff.
MCP is off. AI tools cannot connect until you turn it back on.
No calls in this range.
No calls in this range.
Your agent queries BigQuery through the limits you set.
The desktop app runs an MCP server on your machine. Claude Code, Codex, Cursor, Gemini CLI, Claude Desktop and VS Code connect to it with one button each, and every call they make goes through the caps, budgets, dataset rules and approvals the app already holds you to. The tab above is where those calls land.
A cap and a budget
Maximum GiB per query, plus a daily limit for AI tools alone. A query over either one is refused before BigQuery bills for it. On a reservation, a slot limit per query as well: the query is cancelled once it passes it, and the slot time it used is billed.
On-demand or slots, per query
The same query can cost several times more on one billing model than on the other. Auto routing prices both after the dry run and sends the query to the cheaper one.
Reads run, writes wait
A SELECT inside your limits runs without asking. INSERT, UPDATE, DELETE, MERGE and schema changes are off until you allow them, and then they still stop on a card that shows the statement and the table it changes.
The agent knows its own limits
get_limits reports the caps, today's usage and your dataset rules, so an agent plans a query it is allowed to run instead of finding out by being refused.
One query above the line
When a query needs more than the agent has, it asks you, for that one call, with its reason and the exact limits it would lift. Nothing is saved: the next call is held to your saved limits again.
Or you take it over
Open in a new tab hands the statement to a query tab you drive, on the route the agent asked for. Nothing runs until you run it, and the agent is told you took it.
search_schemaBrowse projects, datasets, tables and columnssample_tableSample rows and column profiles, at no query costvalidate_sqlErrors, warnings and hints before a runestimate_costPrice a query, pick a route, check budgetsget_routingWhere a query would run, and whyget_optimization_tipsWays to make a query read fewer bytesget_bq_docsLook up BigQuery SQL documentationget_querylab_docsRead the Querylab.io documentationexecute_queryRun a query, return rows or write a fileget_job_resultCome back for a query that was still runningcancel_jobAsk BigQuery to stop a running queryget_limitsEvery limit it runs under, with today's usagerequest_limit_increaseAsk you to allow one call above one limitget_approval_resultWait for an answer it is owedThirteen tools, no SQL of ours in between: the agent writes the query, Querylab.io prices it, checks it against your limits and runs it in the project you picked.
Settings → MCP installs the connection, and for Claude Code, Codex and Gemini CLI the instructions that come with it. The server is local: your SQL and your rows do not pass through us, and result rows reach the agent only when you allow it.
On macOS and Windows. The MCP server runs on your machine.
The features you'll use every day.
That's the short list. See all features: lineage, notebooks, import and compare, query history, dbt, the local BigQuery API and more.
The Console manages your cloud.
Querylab.io is built for writing SQL.
Everything Querylab.io does.
The editor, the cost tools, the data tools and the desktop app. Each one links to its docs.
Completions that know the clause
The list matches where the cursor is. A dot after a project, dataset or table lists only the next level.
Real values in WHERE
Type WHERE status = ' and pick from the values that are in the column. No SELECT DISTINCT, nothing billed.
Ghost text without AI
Gray inline text finishes aliases, column lists and join conditions from your schema and your past queries. No model involved. Press Tab.
Checks as you type
Syntax, unknown columns, aggregation mistakes and cost traps get a squiggle and usually a one-key fix.
Pipe syntax
Completions for every |> step and the columns it sees. Convert standard SQL to pipe and back, run any prefix of the chain.
Graph queries
GQL completions from your property graph, JOIN to GRAPH_TABLE and back, results drawn as a graph.
dbt and Dataform
Paste a model and the editor works on the SQL under the template. No CLI, no compile step.
Go to definition, rename
Jump to a CTE, alias or column, list every use, rename it everywhere at once.
Formatting
One shortcut reflows the statement or the selected lines.
Budgets and limits
A byte limit per query, budgets per day, week or month, slot and time limits.
Auto routing
Each query runs on-demand or on your reservation, whichever is expected to cost less.
Tablesample
One comment samples the tables in the query. The status bar shows what the full run would cost.
On-demand or slots
Every query priced both ways, with a note when the other model is clearly cheaper.
AI function costs
BigQuery's estimate leaves out Gemini calls. The status bar says when AI functions bill on top, and a preview shows the prompts, the number of calls and their cost before you run.
LIMIT hints
Tells you when a LIMIT on a clustered table really cuts the bytes read.
Run one CTE
Run or price a query up to one CTE, pipe stage, subquery or statement, from the gutter.
Column lineage
Where each output column comes from, through CTEs, joins, UNNEST and views.
Insights panel
Where the bytes go before you run, and where the time went after.
Execution plan
Each stage of a finished query: rows, slot time, shuffle, and the problems found in it.
Charts in the SQL
A ggsql clause at the end of the query draws a chart that versions with the SQL.
Prism AI
An agent that looks up your schema, dry-runs what it writes and edits the query with your approval.
MCP server
Claude Code, Codex, Cursor and others query BigQuery through the desktop app, within your limits.
AI ghost text
Turn on a model for what the rules can't predict: our cloud model, or one that runs on your device (Chrome, Ollama, LM Studio, Apple Intelligence).
Inline edit
Describe a change to the SQL under the cursor and get it back as a diff.
A table in one click
Schema with a sample value per column, keys, partitioning, and a free preview of the rows.
Query results for free
Turn a result into a local table and query it again in your browser. Nothing is billed and no BigQuery quota is used.
Data compare
Two tables or query results matched by key: the rows added, removed or changed, and in which columns. Two schemas column by column, two queries line by line. Small tables compare in your browser at no cost.
Import
CSV, JSON, Excel, Parquet, Avro, a Google Sheet or the clipboard into a table. The schema is editable first.
Export
Results or a whole table to Sheets, a file, the clipboard, a table or Cloud Storage.
Time travel
Preview a table as it was an hour, a day or a week ago.
Storage costs
What a project, dataset or table costs to keep, and whether the other billing model saves money.
GA4 exports
Which events and parameters are really in the data, from a free sample, with Google's description of each. Click a parameter to insert its typed UNNEST.
Bulk operations
Copy, clone, move, drop or truncate the tables that match a name. You see what happens to each one first, and the SQL opens in a tab for you to run. A drop asks you to type the dataset or project name.
DDL from the tree
Create, alter, copy or drop starts as SQL in a new tab, filled in from the real schema.
Notebooks
Cells that run on their own. Each result is a table the cells below can query by name.
Query history
Every job your projects ran, from any tool, with its cost. Reopen one with its results.
Share links
Send SQL that opens in someone else's Querylab.io. Set an expiry, revoke any time.
Results grid
Sort, filter and page in place, or switch to JSON and job metadata.
Tabs and split view
Tabs keep their SQL, results and limits across reloads. Two tabs fit side by side.
Command palette
⌘K finds tabs, tables and actions.
Notifications
A system notification when a long query or a Prism AI answer finishes.
Local BigQuery API
bq and the Python client run SQL on your computer instead of BigQuery. Nothing is billed.
Your own AI subscription
Prism AI runs through Claude Code, Codex, Gemini CLI, Copilot or Cursor.
gcloud sign-in
Reuse the gcloud credentials already on your machine.
Querylab.io, native on your desktop.
The same IDE without the browser. Your BigQuery credentials and your AI subscriptions stay on your machine.
Apple SiliconWindows 10 and 11|Use the web app instead
Bring your own AI
Prism AI drives the agents already on your machine: Claude Code, Codex, Gemini CLI, Copilot, Cursor. Your subscriptions, your keys.
Read the docs →MCP server
Claude Code, Codex, Cursor and other AI tools query BigQuery through the limits you set.
Read the docs →Existing gcloud auth
Already logged in with gcloud? The app picks it up automatically. No extra sign-in.
Read the docs →Credentials stay local
BigQuery credentials live in your OS keychain, never on our servers.
Read the docs →Native hotkeys
No browser shortcuts getting in the way. No tab to lose.
Read the docs →Local BigQuery API
bq and the Python client run SQL on your computer instead of BigQuery. Nothing is billed.
Read the docs →Learn BigQuery SQL, free in your browser.
76 lessons and 251 challenges on UK car safety tests and a Google Analytics 4 event export. Every answer is checked. No account, no Google Cloud project and no card.
First lesson: SELECT and WHERE, about 12 min. The whole course takes about 14 hours.
A buyer asks how reliable Ford cars are. List every FORD test that failed in June 2024: vehicle_id, completed_date and model.
SELECT vehicle_id
FROM `learn.mot.fact_mot_test`
WHERE make = 'FORD'SQL foundations
Filter, group, join and reshape real car safety check data.
- Basics6 lessons · 70 minFilter, sort and label rows from a table of car safety checks.
- Aggregation7 lessons · 75 minCount, sum and average per group, and stack results with UNION ALL.
- Joins4 lessons · 50 minCombine tables with inner, left and self joins, and find what is missing.
- Dates and strings6 lessons · 70 minWork with dates and text, and convert values safely.
- Subqueries and CTEs3 lessons · 35 minUse one query's result inside another, including recursive CTEs.
- Window functions4 lessons · 45 minRank rows, compare with the previous row and keep the top n per group.
- Arrays and structs5 lessons · 45 minOpen arrays and records, and build lists with ARRAY_AGG.
Advanced BigQuery features
Features that only BigQuery has, such as cost checks and pipe syntax.
- BigQuery for SQL users8 lessons · 75 minHow BigQuery differs: cost, data pruning, wildcard tables and approximate counts.
- Cost and optimization8 lessons · 90 minRead fewer bytes: pick columns, filter partitions, use clustering and read the estimate.
- Pipe syntax3 lessons · 25 minWrite queries as a pipeline that reads in the order it runs.
- Scripting3 lessons · 55 minHold values in variables, branch, loop, and keep one failure from ending the script.
- Writing data7 lessons · 90 minSave results as tables and views, then change rows with INSERT, UPDATE, DELETE and MERGE.
Analytics on real data
Events, sessions, funnels and retention on GA4 export data.
Capstones
An open problem with no step-by-step guide.
You write every answer in Querylab.io, the same editor as the app, with completions, error checks and a cost estimate as you type.
Your data goes from BigQuery to your browser. Nothing sits in between.
Results never reach us
BigQuery returns rows straight to your browser. Our backend never runs your query and never sees the data.
Credentials
Web: OAuth tokens encrypted with AES-256 at rest.
Desktop: your OS keychain.
Revoke anytime from your Google Account.
Queries stay local
Tabs, SQL, history and cached schema live on your device. Cloud sync is off unless you turn it on; shared links are encrypted in your browser.
AI is the exception
Prism AI and AI completions send SQL and table names through our backend to the provider you pick. Off until you use them.
$0
The editor is free.
No credit card and no trial. Sign in with Google and start.
SELECT feature, price
FROM querylab.plans
WHERE plan = 'free'| feature | price | |
|---|---|---|
| 1 | Full SQL editor | $0.00 |
| 2 | Unlimited queries | $0.00 |
| 3 | Cost preview, per CTE | $0.00 |
| 4 | Schema-aware completions | $0.00 |
| 5 | Budget limits, per-query caps | $0.00 |
| 6 | On-demand vs reservation pricing | $0.00 |
| 7 | Prism AI | $0.00 |
| 8 | Pipe syntax | $0.00 |
| 9 | Notebooks | $0.00 |
| 10 | Query history | $0.00 |
| 11 | Lineage | $0.00 |
| 12 | Import, export, data compare | $0.00 |
That's the short list. Everything on the All features tab is free too.
Your BigQuery billGoogle bills you for BigQuery as before. The cost calculator estimates that bill and how much Querylab.io could take off it.Open the calculator →Pipe ↔ SQL
Convert BigQuery pipe syntax to standard SQL and back. Runs entirely in your browser. · Docs →
GQL ↔ SQL
Translate BigQuery graph queries to SQL and back. Runs entirely in your browser. · Docs →
Cost Calculator
What BigQuery will cost you, and what Querylab.io takes off the bill.
Questions? Answers.
Can't find what you're looking for? Reach out.
Why switch from BigQuery Studio?
BigQuery Studio is built to manage your cloud. Querylab.io is built for writing SQL: completions from your real schema, errors caught before you run, and a cost estimate on every query.
Can I try it before committing?
There is nothing to commit to. Querylab.io is free, including commercial use. No credit card, no trial timer, no seat count. The Playground lets you explore without even connecting BigQuery.
What can Querylab.io see?
Not your query results. Queries run from your browser or the desktop app straight to BigQuery with your own Google credentials, and results come straight back. Your SQL and tabs stay on your device unless you turn on cloud sync. On the web, your Google token is stored encrypted on our servers so your session stays signed in. On the desktop, it stays in your Mac's Keychain or the Windows Credential Manager. Prism AI sends SQL and table names to the AI provider, and only when you use it. Security and privacy →
My company blocks third-party Google apps. Can I still use it?
Yes. The desktop app can sign in with your existing gcloud credentials (gcloud auth application-default login). There is no Querylab.io grant for your admin to approve, and the credentials stay on your machine. Use your gcloud sign-in →
Can my own AI tool query BigQuery through Querylab.io?
Yes. The desktop app runs a local MCP server, and Claude Code, Claude Desktop, Codex, Gemini CLI, Cursor and VS Code each connect to it with one button. Every call they make is held to the limits you set: a maximum GiB per query, a daily limit for AI tools, per-dataset rules, and writes that wait for your approval. The MCP activity tab lists every call with what it billed and what it cost. See the MCP tab →
What makes Prism AI better than Cursor with MCP?
You can have both: the desktop app's MCP server lets Cursor or Claude Code query BigQuery through your cost limits. What Prism AI adds is the rest of the work in one window. It searches your real schema before it writes SQL, dry-runs what it writes, puts the price on the Run button, and leaves the result in a grid with history and exports.
What else does Querylab.io do?
Column lineage, notebooks, query history across every tool, data import and compare, dbt and Dataform support, GA4 exports, time travel, and a local BigQuery API on the desktop. The All features tab lists each one with a link to its docs. All features →
How does it reduce BigQuery costs?
Tablesample mode runs test queries on a sample of each table, at a fraction of the cost. BigQuery can't sample views, and a table up to about 168 MiB is one block, so those are read in full. Budget limits block runaway queries. Per-CTE breakdown pinpoints what's expensive. Auto routing sends each query to on-demand or to your slots, whichever its past runs say is cheaper.
How long does it take to get started?
Two minutes. Sign in with Google, authorize BigQuery, start writing. No installation, no config, no onboarding calls. Schema loads automatically.
Is there a desktop app?
Yes, for macOS (Apple Silicon), and in beta for Windows from the Microsoft Store. Credentials stay local in your OS keychain, hotkeys are native, Prism AI drives the agent already on your machine, and the local MCP server lets that agent query BigQuery inside your limits. Download →
Where can I learn BigQuery SQL?
On learn.querylab.io, for free. It has 76 lessons and 251 challenges, from SELECT and joins to window functions, cost control, scripting and GA4 analytics. You write real BigQuery SQL in your browser, and every answer is checked. There is no account, no Google Cloud project and no card. Start learning →
Get in touch.
Questions about BigQuery costs, enterprise pricing, or something that broke? We read every message.
Email us directly
Response time
Usually within 24 hours, Monday to Friday.
Looking for a demo?
Privacy Policy
Last updated: November 3, 2026
This Privacy Policy explains how Querylab.io ("we", "us", "our") collects, uses, stores, and protects your personal data when you use Querylab.io (the "Service"), whether through the web application, the desktop application (together, the "Application"), or our marketing website at querylab.io (the "Landing Site").
1. Data Controller
Querylab.io is the data controller responsible for your personal data. For questions or requests regarding your data, contact us at contact@querylab.io.
2. Information We Collect
2.1 Account Information
- Google OAuth: Email address, display name, and profile picture (collected automatically on first sign-in)
- Product update emails: When you create an account, we add your email address and name to our mailing list, hosted by Mailchimp, so we can tell you about changes to the Service. We send these emails a few times a year. Every one of them has an unsubscribe link, and unsubscribing does not affect your account
- Google Scopes: We request Google permissions incrementally. Only BigQuery access is required to use the Service; all others are optional and requested only when you use specific features:
- BigQuery - Required to run queries (mandatory)
- Cloud Platform (read-only) - Optional, to list your GCP projects
- Cloud Platform (full) - Optional, for scheduled queries and saved queries
- Google Drive (file-level) - Optional, for exporting results to Google Sheets. This scope only grants access to files you explicitly select or create through Querylab.io, not your entire Drive
Google API Services User Data
Querylab.io's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
2.2 Payment Information
- Stripe processes all payments. We store only your Stripe customer ID and subscription status
- We never see or store your credit card number, CVV, or full billing details
2.3 SQL Queries and Data
- Local storage (default): Your SQL queries, tabs, and query history are stored locally in your browser (localStorage and IndexedDB) or desktop app. We do not have access to this data
- Cloud sync (opt-in): If you enable "Store tabs encrypted on backend" in Settings, your tab data (including SQL query text and editor state) is encrypted with AES-256-CBC on our server before it is stored. Query results (the actual data returned by BigQuery) are never stored on our servers
- Shared queries and notebooks: When you share a query or notebook, it is encrypted client-side using AES-256-GCM. The encryption key exists only in the share URL fragment (never sent to our server). Shared notebooks are sanitized before encryption to remove user-specific fields (e.g., billing project). Shares auto-expire after 7 days
- Saved queries: Saved queries are stored in Google Cloud Dataform repositories within your own GCP project, governed by your Google Cloud IAM policies. Google's Dataform API does not accept requests from a web browser, so when you use saved queries in the web application, those requests pass through our backend with your Google access token. We forward them to Google and do not store the query text or the token. The desktop application calls Google directly
- Query results: Query results are returned directly from Google BigQuery to your browser or desktop app. We never see, store, or process your query results
2.4 BigQuery Metadata
- Table names and schemas are cached locally on your device for performance (up to 24 hours)
- This metadata is never sent to our servers. It may be included in Prism AI conversations if you use AI features, in which case it is sent to the AI provider you select (see Section 2.5)
2.5 AI Conversations (Prism AI)
- AI conversations are processed through our backend, which routes requests to the AI provider you select (Anthropic, Google, or OpenAI)
- Conversations are stored locally in your browser (IndexedDB) only. We do not store AI conversations on our servers
- On desktop, you may use CLI providers (e.g., Claude Code) which route conversations directly to the AI provider without passing through our backend. Conversations are still sent to the provider's API
- We use the API tier of each provider, where your data is not used for model training by default. Providers may retain data temporarily for abuse monitoring (typically 30-55 days). See each provider's API data policy: Anthropic, OpenAI, Google
2.6 Analytics
- We collect feature usage and performance metrics (e.g., button clicks, query execution duration, bytes processed) using Google Analytics (GA4) and a custom analytics pipeline
- We use PostHog for anonymous heatmaps (click positions and scroll depth) across the Landing Site and Application. Heatmap data contains no personally identifiable information
- On the Landing Site, PostHog session replay is enabled for visitors who accept cookie consent. This records anonymous interactions (clicks, navigation, screen content) to help us improve the experience. No session replay is recorded for visitors who decline consent
- Query cost data: When a query finishes, we record the resources it used in BigQuery (bytes processed, slot time), how it was billed (region, on-demand or reservation, edition, estimated prices), and numbers that describe its structure: its length, counts such as the number of joins, subqueries and window functions, and which BigQuery built-in functions it calls. We use this to improve how Querylab.io chooses between on-demand and reservation pricing. The record never contains your SQL text, table, column, dataset or project names, or the values in your query. Runs of the same query on one device share an identifier computed with a random key that never leaves that device, so without the key the identifier cannot be matched to a query
- We never collect: SQL query text, table names, column names, query results, or any of your BigQuery data in analytics
- Learn site: learn.querylab.io sends anonymous usage events, such as which lessons are started and how far you got, and anonymous crash reports that never include the error message. The first event of a visit records the domain that linked to it and the
utm_tags of the link, when it had them. Without your consent, it sets no cookies, takes no location from your IP address, and sends none of your queries or data. It sends nothing when your browser sends the Global Privacy Control signal - Learn site, with your consent: visitors in the EU/EEA, UK and Brazil are asked on their first visit; visitors elsewhere are not asked. If you accept or are not asked, Learn also loads Google Analytics, with all advertising features denied, and PostHog records page views and anonymous session replays. Replays show the lesson text and where you move, click and scroll. They hide your SQL, your results, error messages, your account name and photo, and anything you type. To decide whether to ask you, your browser makes one request to ipapi.co. You can withdraw your consent at any time in Settings
- Querylab.io Local: local.querylab.io sends the same anonymous events as the Learn site, without the ones about lessons. It loads no Google Analytics and records no replays. Cloudflare counts its page views like the other sites (see Page view counts below)
- Page view counts: The Landing Site, the docs site (docs.querylab.io), the Learn site and Querylab.io Local use Cloudflare Web Analytics to count page views, the site that linked to the page, and the visitor's country, browser and device type. It sets no cookies, stores nothing in your browser and does not identify or follow you across visits, so it runs without consent. The Application does not use it
- Cookie consent: Users in the EU/EEA, UK, and Brazil are shown a cookie consent banner (see Section 9 for details)
- Opt-out: Logged-in users can disable analytics entirely in Settings (this fully stops all analytics collection). Landing Site visitors can decline persistent tracking via the cookie consent banner
2.7 Playground Mode (No Account Required)
- Playground users are identified by an anonymous ID. We do not collect personal information or use browser fingerprinting for playground users
- Playground queries are executed on a shared BigQuery project managed by us. Unlike logged-in users (whose queries go directly to their own Google Cloud), playground queries pass through our backend infrastructure
- We may log playground queries for abuse prevention, service improvement, and enforcing fair-use limits. We recommend that you avoid entering sensitive or proprietary information in Playground mode
- We track aggregate usage (bytes processed, query count) to enforce fair-use limits
- Session replay: We use PostHog to record anonymous session replays of playground interactions (UI clicks, navigation, screen content). This helps us understand how users interact with the product. Session replays are not linked to any personal identity. No cookies are used — data is stored in memory only for the duration of the session. Session replay is not enabled for logged-in users in the Application
- Playground data is automatically deleted after a period of inactivity
3. Legal Basis for Processing (GDPR)
| Processing Activity | Legal Basis |
|---|---|
| Account creation and authentication | Contract performance (Art. 6(1)(b)) |
| Payment processing | Contract performance (Art. 6(1)(b)) |
| Cloud sync of tabs (opt-in) | Consent (Art. 6(1)(a)) |
| AI conversation processing | Contract performance (Art. 6(1)(b)) |
| Analytics and product improvement (EU/EEA/UK/Brazil) | Consent (Art. 6(1)(a)) — via cookie consent banner |
| Analytics and product improvement (other regions) | Legitimate interest (Art. 6(1)(f)) |
| Product update emails to account holders | Legitimate interest (Art. 6(1)(f)) — you can object at any time with the unsubscribe link in every email |
| Ad conversion measurement (Reddit Ads) | Legitimate interest (Art. 6(1)(f)) |
| Security and abuse prevention | Legitimate interest (Art. 6(1)(f)) |
4. Data Storage and Security
- Desktop app: Refresh tokens and sensitive credentials are stored in your OS keychain (macOS Keychain, Windows Credential Manager, or Linux Secret Service). Your Google access and refresh tokens never leave your device. When you sign in with local gcloud credentials, the application requests short-lived tokens from gcloud and stores none
- Web app: Your Google refresh token is encrypted with AES-256 and stored on our backend, so your session can be renewed without signing in again. We use it only to renew access tokens for your own session and to refresh your Google profile (email address, name and picture). Our backend does not use it to call BigQuery. Querylab.io session tokens are stored in HTTP-only secure cookies with rotation
- Backend: Data is stored in Google Cloud Platform (Firestore). All data is encrypted at rest by Google Cloud and in transit via TLS
- Cloud sync: Tab data is encrypted with AES-256-CBC using a server-side key before storage
- Shared queries and notebooks: End-to-end encrypted with AES-256-GCM. The decryption key is in the URL fragment and never reaches our servers
- Payments: Processed by Stripe (PCI DSS Level 1 certified)
5. International Data Transfers
Your data may be processed in the United States through our use of Google Cloud Platform, Stripe, Mailchimp, and AI providers (Anthropic, Google, OpenAI). These transfers are protected by:
- EU-US Data Privacy Framework (where applicable)
- Standard Contractual Clauses (SCCs) as approved by the European Commission
- The service providers' own GDPR compliance commitments
6. Third-Party Services
| Service | Purpose | Data Shared |
|---|---|---|
| Google Cloud Platform | Authentication, BigQuery access, data storage | Account info, encrypted tab data (if opt-in) |
| Stripe | Payment processing | Billing info (managed by Stripe) |
| Anthropic | AI assistance (Prism AI) | Conversation content (when selected as provider) |
| Google AI (Gemini) | AI assistance (Prism AI) | Conversation content (when selected as provider) |
| OpenAI | AI assistance (Prism AI) | Conversation content (when selected as provider) |
| Google Analytics (GA4) | Product analytics | Anonymous usage metrics (no PII). Cookieless mode for users who decline consent. Also on the Learn site, only for visitors who accept or are not asked (see Section 2.6) |
| PostHog | Heatmaps, session replay, Landing Site analytics | Anonymous click positions, scroll depth, pageviews. Session recordings for playground users in the Application (memory-only) and consenting visitors on the Landing Site (localStorage). On the Learn site, anonymous usage events for everyone, and page views and masked session recordings for visitors who accept or are not asked. No PII |
| Cloudflare Web Analytics | Page view counts on the Landing Site, the docs site, the Learn site and Querylab.io Local | The page address, the address of the page that linked to it, and your browser's user agent. Your IP address is used only to find your country and is not stored. No cookies |
| ipapi.co | Finding your country, to decide whether to show the cookie consent notice (Landing Site and Learn site) | Your IP address, sent from your browser. We do not receive the answer; your browser keeps a yes or no for 7 days |
| Mailchimp (Intuit) | Product update emails | Email address, name, the IP address and date of sign-up (kept as the record of when the address was added), and the sign-up source (web app, desktop app, or landing page form) |
| Reddit Ads (Conversions API) | Ad conversion measurement (server-side only) | Hashed email and hashed IP address, user agent. Sent server-to-server on signup and purchase events only — no client-side tracking pixel or cookies |
7. Data Retention
- Account data: Retained while your account is active. Deleted within 30 days of account deletion request
- Payment records: Retained for 7 years per Belgian tax law
- Mailing list: Your address is removed from the list when you delete your account, when you unsubscribe, or when you ask us at contact@querylab.io. Mailchimp keeps a record of unsubscribed addresses so that they are not added again
- Shared queries and notebooks: Auto-expire and are deleted after 7 days
- Analytics data: Anonymized analytics data is retained without a fixed expiration period, as it contains no personal data (see Section 2.6)
- Playground usage data: Automatically deleted after inactivity
- AI conversations: Stored locally in your browser only. Clearing browser data removes all conversations
8. Your Rights
Under the GDPR and LGPD (Brazil), you have the right to:
- Access: Request a copy of all personal data we hold about you
- Rectification: Correct inaccurate personal data
- Erasure: Request deletion of your personal data ("right to be forgotten")
- Data portability: Receive your data in a structured, machine-readable format
- Restriction: Request restriction of processing in certain circumstances
- Objection: Object to processing based on legitimate interests, including analytics and product update emails (use the unsubscribe link in any email)
- Withdraw consent: Where processing is based on consent (e.g., cloud sync), you may withdraw at any time
To exercise these rights, contact contact@querylab.io. We will respond within 30 days as required by GDPR.
9. Cookies
- Essential cookies: Used for authentication and session management. These are strictly necessary and do not require consent
- Analytics cookies: Google Analytics (GA4) may set cookies for usage tracking. On the Landing Site and the Learn site, PostHog uses localStorage (not cookies) for analytics persistence when consent is granted or not required. In the Application, PostHog always uses memory-only persistence (no cookies or localStorage)
- Cookie consent banner: Users in the EU/EEA, UK, and Brazil are shown a consent banner on first visit. Accepting enables persistent identifiers (cookies and localStorage) for cross-session analytics. Declining disables persistent identifiers — anonymous, aggregate analytics data is still collected but cannot be linked across visits. Your choice is stored in localStorage and respected on future visits
- Learn site: Learn asks for consent the same way as the Landing Site. If you decline, it stores no analytics identifier and loads no Google Analytics or session replay. Withdrawing consent in Settings removes the identifier and the Google Analytics cookies
- Other regions: Persistent analytics identifiers are enabled by default. Logged-in users can disable all analytics entirely in Settings
- We do not use advertising cookies or third-party tracking cookies
10. Children's Privacy
Querylab.io is not intended for users under 16 years of age. We do not knowingly collect personal data from children. If you believe a child has provided us with personal data, contact us and we will delete it.
11. Data Breach Notification
In the event of a personal data breach that poses a risk to your rights and freedoms, we will notify the relevant supervisory authority within 72 hours and inform affected users without undue delay, as required by applicable data protection laws including GDPR (Articles 33 and 34) and LGPD (Article 48).
12. Changes to This Policy
We may update this Privacy Policy. Material changes will be announced via email at least 30 days in advance. Where required by law, we will obtain your consent before applying changes that affect consent-based processing.
13. Contact
For privacy inquiries or to exercise your rights:
Email: contact@querylab.io
You also have the right to lodge a complaint with the Belgian Data Protection Authority (APD/GBA) at www.dataprotectionauthority.be.
Terms of Service
Last updated: April 30, 2026
These Terms of Service ("Terms") govern your use of Querylab.io (the "Service"). By accessing or using the Service, you agree to these Terms. If you do not agree, do not use the Service.
1. Definitions
- "Service" means the Querylab.io web application, desktop application, and all related services
- "User", "you" means any individual or entity using the Service
- "Content" means SQL queries, data, and other materials you create or upload through the Service
- "Prism AI" means the AI-powered assistant features within the Service
- "Playground" means the limited-access mode available without registration
2. The Service
Querylab.io is a SQL IDE for Google BigQuery with code editing, query execution, cost analysis, and AI assistance. The Service connects to your Google Cloud resources but does not execute queries on its own infrastructure - all query execution occurs within your Google Cloud account.
We depend on third-party services including Google BigQuery, Google Cloud Platform, Stripe, and AI providers (Anthropic, Google, OpenAI). We are not liable for interruptions or changes to these third-party services.
3. Account Registration
- You must authenticate via Google OAuth to create an account
- You are responsible for all activity under your account
- Querylab.io is intended for users 16 years or older
- You must not share your account credentials with others
4. Playground Mode
The Playground provides limited access to the Service without registration. By using the Playground, you agree to these Terms and our Privacy Policy. Playground usage is subject to fair-use limits (currently 5 GiB per query, 50 GiB per day) and may be restricted or discontinued at our discretion.
Playground queries are executed on shared infrastructure managed by us and are not private. You should not enter sensitive, confidential, or proprietary data in Playground mode. See our Privacy Policy for details on how playground data is handled.
5. Plans and Billing
| Plan | Details |
|---|---|
| Free | Free for individual and commercial use. Includes core SQL IDE features and standard Prism AI quota |
| Teams | Org governance, premium AI quota, and team features. Monthly or annual billing. New business users receive a 30-day free trial (no payment required; automatically converts to Free plan if not upgraded) |
| Enterprise | Seat-based licensing for organizations with custom terms, deployment in your own Google Cloud project, and SLAs |
Stripe processes all payments. Subscriptions auto-renew until cancelled. You may cancel at any time in Settings. See our Refund Policy for cancellation and refund terms.
6. BigQuery Costs
You are solely responsible for all Google BigQuery costs incurred through Querylab.io. While we display cost estimates before query execution, these are approximate and not guarantees. Actual billing is determined by Google. We are not liable for unexpected charges. Report significant estimate-to-actual variances to us at contact@querylab.io.
7. Acceptable Use
You agree not to:
- Use the Service for any illegal purpose
- Attempt unauthorized access to our systems, other accounts, or Google Cloud resources you do not own
- Interfere with or disrupt the Service or its infrastructure
- Use automated systems that exceed reasonable usage limits (100 queries per minute)
- Reverse engineer, decompile, or attempt to extract the source code of the Service
- Resell, redistribute, or sublicense access to the Service without our written consent
- Use the Service in violation of Google Cloud or BigQuery terms of service
You represent that you have proper authorization to access all BigQuery resources you query through Querylab.io.
8. AI Features and Data
- Prism AI may process your queries and conversation content via third-party AI providers (Anthropic, Google, or OpenAI). You choose which provider to use
- We access BigQuery metadata (table names, schemas) for intelligent features such as auto-completion and diagnostics. We do not access or store your actual BigQuery data or query results
- AI-generated suggestions (SQL rewrites, explanations, optimizations) are provided as-is. You are responsible for reviewing and validating any AI output before executing it
- See our Privacy Policy for details on data handling
9. Free Tools
Querylab.io may offer free online tools (such as the SQL Pipe Syntax Converter) accessible without registration. These tools are subject to all provisions of these Terms, including but not limited to:
- The underlying software, algorithms, parsers, and code powering these tools are proprietary intellectual property of Querylab.io
- You may not copy, extract, download, decompile, reverse engineer, or redistribute the JavaScript bundles, source code, or any derived works from these tools
- You may not use automated scripts, bots, or scraping tools to interact with these tools beyond normal individual use
- Individual and commercial end-use of these tools is permitted; embedding them in your own product, reselling access, or integrating the underlying code into other products requires our written consent
- We may modify, rate-limit, or discontinue any free tool at any time without notice
Violation of these terms may result in access restrictions and legal action.
10. Intellectual Property
- Our IP: Querylab.io, its software, design, and documentation are our intellectual property. These Terms grant you a limited, non-exclusive, revocable license to use the Service
- Your Content: You retain all rights to your SQL queries, data, and other content. You can export your data at any time via account settings
11. Disclaimer of Warranties
THE SERVICE IS PROVIDED "AS IS" AND "AS AVAILABLE" WITHOUT WARRANTIES OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT. WE DO NOT WARRANT THAT THE SERVICE WILL BE UNINTERRUPTED, ERROR-FREE, OR THAT COST ESTIMATES WILL BE ACCURATE.
12. Limitation of Liability
TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW:
- We are not liable for indirect, incidental, special, consequential, or punitive damages, including lost profits, data loss, or business interruption
- Our total aggregate liability for any claim arising from or relating to the Service is limited to the amount you paid us in the 12 months preceding the claim, or EUR 100, whichever is greater
- This limitation does not apply to damages caused by our gross negligence or willful misconduct, or to any liability that cannot be excluded under applicable law
Nothing in these Terms limits your rights as a consumer under Belgian or EU consumer protection law.
13. Indemnification
You agree to indemnify and hold Querylab.io harmless from any claims, damages, or expenses arising from your violation of these Terms, misuse of the Service, or unauthorized access to BigQuery resources through the Service. This does not apply to claims arising from our gross negligence.
14. Termination
- By you: You may stop using the Service and delete your account at any time. Your data will be handled per our Privacy Policy
- By us: We may suspend or terminate your account for violation of these Terms, with reasonable notice where possible. For paid plans terminated by us without cause, we will provide 30 days notice and a pro-rata refund
- Sections 6, 9, 10, 11, 12, 13, and 15 survive termination
15. Governing Law and Disputes
These Terms are governed by Belgian law. Any disputes shall first be resolved through good-faith negotiation. If negotiation fails, disputes shall be submitted to the courts of Belgium. EU consumers retain the right to bring claims in their country of residence. You may also use the EU Online Dispute Resolution platform at ec.europa.eu/consumers/odr.
16. General Provisions
- Entire agreement: These Terms, together with the Privacy Policy and Refund Policy, constitute the entire agreement between you and Querylab.io
- Severability: If any provision of these Terms is found invalid or unenforceable, the remaining provisions continue in full force
- No waiver: Failure to enforce any provision does not constitute a waiver of that provision
- Assignment: You may not assign your rights under these Terms. We may assign our rights in connection with a merger, acquisition, or sale of assets
- Force majeure: We are not liable for failures or delays caused by circumstances beyond our reasonable control, including natural disasters, wars, pandemics, government actions, or third-party service outages
17. Independence
Querylab.io is an independent product and is not affiliated with, endorsed by, or sponsored by Google LLC, Anthropic, OpenAI, or any other third party.
18. Changes to These Terms
We may update these Terms. Material changes will be announced via email at least 30 days in advance. Continued use of the Service after changes take effect constitutes acceptance. If you do not agree to the updated Terms, you must stop using the Service.
19. Contact
Email: contact@querylab.io
