QRY
QRY is an enterprise platform that lets teams query their databases in plain language. Its AI Data Analyst interprets the question, writes the SQL automatically and returns charted answers across eight native connectors and fifty more through Trino.
What is QRY?
QRY calls itself a Corporate Intelligence Platform and is explicit that it is not another BI tool: rather than replacing the data stack, it sits as a transversal layer on top of whatever already exists. The component that carries the product name is the AI Data Analyst, and its job is narrow and well defined. You ask a question in ordinary language, it reads your question against the database schema, writes the SQL itself, runs it and returns the answer in a readable form. The generated SQL is never hidden: it is displayed and can be edited, which the documentation presents both as a safeguard and as a way to learn.
Reach is the main selling point. Eight databases are supported natively — PostgreSQL, Snowflake, BigQuery, Databricks, Starburst, Redshift, Cloudera and Salesforce — and more than fifty others become reachable through an integration with Trino, including MySQL, Oracle, SQL Server, MongoDB, Cassandra, Redis and Elasticsearch. A single conversation can span several of them. Files can also be uploaded directly in CSV, Excel, JSON or Parquet format when there is no database to connect.
The model layer is deliberately open. Administrators choose which engines their users may call, from commercial ones such as Claude, GPT and Gemini to open models like Llama, Qwen and DeepSeek run locally through Nvidia NIM — the option that makes air-gapped deployment possible. To get past the memory limits of language models, QRY generates and executes Python with matplotlib and supports out-of-core processing and Spark, which is how it claims to analyse very large datasets.
Around that core sit the features an enterprise buyer will look for: persistent memory across conversations with semantic search over history, a domain context system that ingests internal documentation so answers use company vocabulary, domain agents summoned with an @ mention, plain-language scheduling of recurring reports, hierarchical role-based access control down to table level, attribute-based row-level security, PII detection and masking, audit trails, and per-user or per-team budget caps on model spending. Notebooks, dashboards, workspaces, data products and an ETL module extend it further. Three deployment modes are offered: managed SaaS, private cloud in the customer's own VPC, and fully on-premise. Fifty or more query languages are claimed.
What it does
- Answer a business question asked in plain language with figures pulled straight from the database
- Generate the SQL automatically, then show it and let you edit it
- Query several databases at once inside a single conversation
- Run Python and produce matplotlib charts on datasets too large for a model's context window
- Schedule recurring reports described in plain language, with no cron job to write
- Analyse an uploaded CSV, Excel, JSON or Parquet file without connecting a database at all
- Enforce row-level permissions, mask personal data and log every access for audit
When to use QRY / When not to
A quick filter to help you decide if QRY is the right fit.
When to use QRY
- Data analysts buried under ad hoc reporting requests who want to automate the repetitive part
- Business teams who need figures without waiting on a ticket for the data team
- CTOs, CIOs and chief data officers looking to open up data access without loosening governance
- International organisations whose staff query the same data in different languages
- Regulated or sovereignty-conscious companies needing on-premise or air-gapped deployment with local models
When not to use QRY
- Individuals and small businesses: the published entry point is 10,000 USD per year
- Anyone wanting to sign up and test immediately, since every button leads to a contact form
- Teams with no database or data warehouse to connect, as the tool has nothing to query
- Mobile-first users, because there is no iOS or Android application
- People looking for a general-purpose assistant or a content generator rather than a data tool
How to use QRY
A typical end-to-end flow, from setup to results.
- Contact the vendor through the form on the site: there is no self-service sign-up
- Choose a deployment mode, either managed SaaS, private cloud in your own VPC, or on-premise
- Connect your data sources using the eight native connectors, or reach the rest through Trino
- Alternatively, upload a CSV, Excel, JSON or Parquet file to start analysing without a database
- Have an administrator select which AI models are available and set the default per user group
- Configure role-based and attribute-based permissions, then set budget caps and model quotas
- Feed the domain context system with your internal documentation so answers use your vocabulary
- Ask your question in plain language in the conversational interface
- Read the generated SQL, edit it if it needs adjusting, and ask follow-up questions in context
- Schedule the queries worth repeating, or call the REST API and stream the answers over SSE
Pros & Cons
Pros
- Removes the round trip to the data team for everyday ad hoc questions
- The generated SQL is shown and editable, so the tool is not a black box
- Very wide source coverage through Trino, including legacy systems a modern stack ignores
- On-premise and air-gapped deployment with local models answers real sovereignty constraints
- Governance is built in rather than bolted on, down to row-level rules and audit trails
- Model spending is capped and tracked per user or team, which is rare at this level
- A documented REST API with streaming responses, and public product documentation
Cons
- There is no pricing page at all; the only public figure sits in a comparison table
- Every route, including the free trial and live demo buttons, ends at a contact form
- SOC 2 Type II is shown as in progress and ISO 27001 as pending, so neither is actually held
- No data processing agreement is published, and no subprocessor list exists
- Nothing is said about whether customer data trains models, or how to opt out if it does
- No hosting country is named anywhere, only multi-region deployment and residency controls
- Performance and ROI figures are striking but unsourced, and the case studies name no customer
Pricing & Plans
No permanent free plan is documented, and the site publishes no pricing page. The only public figure appears in the comparison table on the platform page, which places QRY at from 10,000 USD per year. Prices for the named tiers are not disclosed, and the terms indicate that subscriptions are billed in advance monthly or annually, that usage-based charges are billed monthly, and that large accounts may negotiate bespoke arrangements. A free trial is advertised but is obtained by contacting the vendor rather than by signing up.
- Starter — 99.9% uptime SLA
- price not published
- Professional — 99.9% uptime SLA
- price not published
- Enterprise — 99.99% uptime SLA
- price not published
Data, GDPR & hosting
A consolidated view of how QRY handles your data.
GDPR overview
The site claims GDPR compliance explicitly: a compliance badge on the homepage reads GDPR — Compliant, and the privacy policy states full compliance with EU data protection rules alongside CCPA and PIPEDA. Five data subject rights are listed by name: access, rectification, erasure, portability and objection. The publisher is established in Madrid, so no Article 27 representative is required or named, and it designates a data protection officer reachable at protecciondedatos@ixen.ai; the Spanish authority AEPD is cited as the supervisory body. The privacy policy is dated 25 December 2024 and the terms took effect on 1 January 2025. Concrete gaps remain: no data processing agreement is published or offered, no subprocessor list exists, and no hosting country is named.
Who owns the data?
The terms of service state that the customer keeps ownership of both the data connected to the platform and the results of the queries run against it, while the vendor keeps ownership of the platform, its algorithms and any improvements made to them. The customer grants a limited licence allowing the vendor to process that data for the sole purpose of delivering the service. Beyond that, the vendor reserves the right to use aggregated and anonymised data to improve the service. The privacy policy adds that personal data is never sold, and that database connection metadata is collected rather than the underlying business records. Sharing with third parties is limited to consent, legal obligation, vetted providers bound by confidentiality, and corporate transactions such as a merger.
Reuse rights
Because customers keep ownership of their data and of the query results, they may reuse both freely and without asking permission. The SQL that the AI generates is displayed and can be edited, so the work produced inside the tool can be taken away and reused elsewhere. On termination, account data stays available for thirty days so it can be exported. In exchange the terms place clear duties on the customer: only connect databases you are authorised to connect, maintain your own access controls, comply with the privacy rules that apply to you, and keep independent backups. Reverse engineering the platform, extracting its algorithms, sharing credentials or probing systems you have no right to reach are all prohibited.
Data retention & training
Hosting summary
The site names no hosting country and no specific region. It states that deployment is multi-region with data residency controls, and mentions regional data residency among its network optimisations, but never says where data physically sits. In practice the answer depends on the deployment mode chosen. Under managed SaaS the vendor hosts; under private cloud the workload runs in the customer's own VPC; and on-premise, including air-gapped installations, everything stays on the customer's infrastructure, which is the configuration that gives the strongest residency guarantee. The publisher is established in Madrid, Spain, and its corporate privacy policy states that international transfers rely on the EU-US Data Privacy Framework. The marketing site's own domain resolves to Google infrastructure on an anycast address, which describes the website rather than the product. Any buyer with residency obligations should obtain the commitment in the contract, since it cannot be verified from the site.
Things to keep in mind
Risks and trade-offs to weigh before adopting QRY.
- A confident answer in plain language hides the assumptions behind it: if the generated SQL is never read, a wrong join or a misread column becomes a business decision nobody checked
- Making data trivially easy to query invites people to stop learning the underlying model, and the ability to challenge a surprising number erodes with it
- The free trial and live demo buttons do not open a product but a contact form, so budget expectations should be set before internal announcements are made
- Certification claims are inconsistent between pages: the homepage marks SOC 2 Type II as in progress and ISO 27001 as pending, while other pages state compliance in the present tense
- Nothing published says whether queries or results are used to train models, and no opt-out is documented, which matters because prompts can be routed to third-party providers
- No hosting country is named, so a buyer with data residency obligations cannot verify them from the site and must obtain the commitment contractually
- The terms invoke Delaware law and Delaware courts while the publisher is a Spanish company, which could complicate any dispute and should be clarified before signing
Setup & Integrations
Technical difficulty
Two very different levels. Deployment is an IT project: sources must be connected, roles and row-level rules configured, budget quotas set and the domain context populated, and an on-premise install on Kubernetes needs real infrastructure skills. Published case studies quote twenty-one to forty-five days. Managed SaaS is considerably lighter than an air-gapped deployment. Daily use, by contrast, requires nothing technical: questions are asked in ordinary language and the site claims minutes to productivity against weeks or months for traditional BI tools. There is no self-service onboarding, so setup always begins with the vendor.
Deployment
Integrations
Supported languages
Behind QRY
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement QRY.
Frequently asked questions
What does QRY actually do?
Which databases can it connect to?
Do I need to know SQL to use it?
Which AI models does it use?
Can it be installed on our own infrastructure?
How much does it cost?
Is there a free trial?
Is my data used to train AI models?
Who is behind QRY?
How long does deployment take?
Should you pick QRY?
QRY is a serious piece of enterprise engineering wrapped in a thin commercial shopfront. The product itself is coherent and unusually complete for its category: a conversational analyst that writes visible, editable SQL; federation over eight native databases and fifty more through Trino; Python execution to get past model memory limits; and a governance layer with row-level rules, PII masking, audit trails and budget caps that reads as though it were designed with a CISO in the room rather than added later. The ability to run open models locally for air-gapped deployment is a genuine differentiator for regulated buyers, and the public documentation and documented REST API suggest real technical maturity.
The reservations are almost all commercial rather than technical. There is no pricing page; the single public figure is a line in a comparison table. Every call to action, including the ones promising a free trial or a live demo, opens the same contact form. The security posture is announced more confidently on some pages than the certification status supports, with SOC 2 Type II still in progress and ISO 27001 pending. No data processing agreement, no subprocessor list and no named hosting country are published, and the site is silent on whether customer data is ever used to train models. The headline performance and ROI numbers are striking but unsourced, and the case studies name no customer.
The result suits an IT or data organisation able to run a proper evaluation, ask the awkward questions and negotiate a contract. It is not a tool an individual or a small team can try on a whim. Buyers should also note that the legal entity behind the product, PUEDATA S.L. in Madrid, appears nowhere on the QRY site, and that the terms invoke Delaware law despite the publisher being Spanish.
- Choosing a selection results in a full page refresh.
- Opens in a new window.