Qrly
Qrly is a self-hostable business intelligence platform with AI built in. Ask questions in plain English and get SQL back, build dashboards and OLAP models, and bring your own language model, cloud or fully on-premise.
What is Qrly?
Qrly is a self-hostable business intelligence platform whose AI layer is part of the product rather than a paid add-on. It is published by a Belgian company and presents itself as made in Belgium, with the product itself dated 2026. Everything runs on infrastructure the customer controls: Java 25 on Spring Boot 4 with a PostgreSQL application database, capable of a GraalVM native image, deployable on bare metal, a virtual machine or Kubernetes.
The AI side centres on Ask, which turns a natural-language question into QQL and then into SQL. Conversations are multi-turn and schema-aware, a chart type is suggested, and nothing destructive ever runs automatically — a human reviews the statement first. The model is chosen by the customer: Anthropic Claude, Google Gemini, OpenAI, Azure OpenAI and Mistral are supported alongside local runtimes such as Ollama, LM Studio, Jan.ai and LocalAI, so sensitive schemas can stay inside the network. Keys are encrypted per organisation and capped by daily token budgets. Beyond Ask, the AI writes schema documentation, analyses query performance, suggests filters, and scans verified questions for anomalies that surface in an Insights feed. A separate agent layer offers four personas — Analyst, Composer, Modeler and Investigator — running 29 tools on which every write stages a proposal for human approval.
The BI side is conventional in the best sense. QQL, a single JSON query document, compiles to twelve SQL dialects across forty connection types. There are OLAP star and snowflake models with rollup, cube and pivot, thirteen chart types on a twelve-column dashboard grid, four opt-in cache tiers ending in change-data-capture table sync through Debezium, and a spreadsheet engine of roughly five hundred worksheet functions whose cells bind to live SQL. Multi-tenancy runs from tenant to organisation to project, identity combines Azure AD, Google, LDAP and Active Directory simultaneously, and a documented Data API plus a read-only MCP server expose the whole thing programmatically.
Pricing is unusual: an annual fee on the customer's own revenue, in marginal tranches, with every feature shipped at every tier.
What it does
- Ask a question in plain English and get compiled QQL or SQL back for review before it runs
- Build questions in a visual builder or in raw SQL, and convert between the two in either direction
- Publish live dashboards that refresh over WebSocket or PostgreSQL LISTEN/NOTIFY
- Detect anomalies automatically and promote any finding into a saved alert
- Schedule reports to email, Slack, Microsoft Teams, webhook or SMS on a cron cadence
- Embed a branded analytics portal in your own product with signed JWTs and locked parameters
- Let AI agents propose dashboards or OLAP models that a human approves before anything is written
When to use Qrly / When not to
A quick filter to help you decide if Qrly is the right fit.
When to use Qrly
- Data and analytics engineers who want a SQL editor, OLAP star and snowflake models, query lineage and an AI performance analyzer in one place
- Business and operations analysts who need answers without writing SQL, through a visual builder and natural-language questions
- SaaS vendors embedding analytics for their own customers, using signed-JWT embeds with locked parameters and unlimited external viewers
- Managed service providers and consultancies serving several clients from one installation, with per-tenant branding, domains and language-model keys
- Regulated organisations — the vendor names banks, defence and the public sector — that must keep schemas and data inside their own network
When not to use Qrly
- Teams wanting a ready-to-use hosted service: there is no vendor cloud at all, and someone must run Java, Spring Boot and PostgreSQL
- Buyers who require the vendor to hold SOC 2 or ISO 27001, since the publisher states plainly that it holds no certification of any kind
- Organisations that mandate native multi-factor authentication or row-level security, as neither is implemented in the product
- Anyone who wants to evaluate the tool alone before talking to sales: there is no free plan, no free trial and no public demo
- Buyers who need standard published contractual documents, because no terms of service, privacy policy or data processing agreement exists on the site
How to use Qrly
A typical end-to-end flow, from setup to results.
- Contact the publisher through the form on the home page — there is no self-service sign-up
- Provision a server and a PostgreSQL 14+ database on your own infrastructure or private cloud
- Install the application on Java 25 and Spring Boot 4, on bare metal, a virtual machine or Kubernetes
- Run the eight-step first-run wizard, which probes the database read-only and writes nothing until you confirm
- Create the first tenant, organisation and superuser at the confirmation step, then install the systemd unit
- Declare a JDBC connection to your data source and run schema sync to introspect tables and columns
- Configure a language-model provider with your own API key, a token budget and, if needed, a local endpoint
- Ask a question in plain English, or build one in the visual builder or the SQL editor, then review and save it
- Assemble saved questions into a dashboard and set filters, live refresh and permissions
- Add alerts, scheduled subscriptions or a signed embed token to deliver the result where it is needed
Pros & Cons
Pros
- Every feature ships at every price tier — nothing is gated behind a higher plan
- Pricing is indexed on the customer's revenue, so adding users, viewers or queries never raises the bill
- The full price schedule is public, from a €180 floor to a €93,750 ceiling, with the marginal rates published
- No margin is taken on AI: the customer pays their model provider directly, and a local model costs only electricity
- Genuine data sovereignty is achievable — local language models, no phone-home and no third-party asset requests
- The security page publishes its own gaps, naming missing MFA, missing row-level security and the absence of certification
- The audit log records the verbatim SQL executed rather than a summary, alongside row count, duration and status
Cons
- There is no hosted option: the customer must install and operate a Java and PostgreSQL service themselves
- The publisher holds no compliance certification at all, and says so explicitly
- Multi-factor authentication and row-level security are not implemented in the product
- No terms of service, privacy policy, legal notice or data processing agreement is published anywhere on the site
- There is no free plan, no free trial and no public demo, so evaluation requires contacting sales
- The publishing company is named on no page of the site — only a postal address and two phone numbers identify it
- The product is very new, with no named customer, no testimonial and no archived history of the site
Pricing & Plans
There is no permanent free plan and no free trial. Qrly is licensed as an annual subscription whose amount is a percentage of the customer's own annual revenue, charged in marginal tranches. The lowest price point is €180 per year, excluding VAT, which is what a business with €100,000 of annual revenue pays; this is stated as an absolute floor. Billing is annual and in euros only, and covers one organisation and one installation, on-premise or in a private cloud. Users, dashboards, queries and embedded viewers are never metered, and AI tokens are paid directly to the model provider.
- Micro — €180 per year
- for €100k of annual revenue
- Starter — €360 per year
- for €200k of annual revenue
- Growth — €900 per year
- for €500k of annual revenue
- Business — €1
- 500 per year
- for €1M of annual revenue
- Scale — €3
- 600 per year
- for €5M of annual revenue
- Scale+ — €9
- 000 per year
- for €25M of annual revenue
- Enterprise — €21
- 750 per year
- for €100M of annual revenue
- Group — €63
- 750 per year
- for €500M of annual revenue
- Global — €93
- 750 per year
- for €1B of annual revenue
- All tiers include every feature
- they are the same marginal-tranche schedule evaluated at round revenue figures
Data, GDPR & hosting
A consolidated view of how Qrly handles your data.
GDPR overview
The site makes no GDPR statement at all: the words GDPR and data protection regulation appear nowhere across the pages collected, and there is no privacy policy, no legal notice and no data processing agreement to consult. This is an explicit absence rather than a refusal — the publisher neither claims compliance nor denies it. What the publisher offers instead is an architectural argument. Qrly is installed on the customer's own infrastructure, so personal data never reaches the vendor, there is no sub-processor to declare and data residency follows the customer's own hosting choice. The publisher is established in Belgium and therefore falls within the European Union itself, which is why no Article 27 representative is designated. Buyers who need documented GDPR commitments will have to obtain them contractually, because the website provides none.
Who owns the data?
Qrly runs entirely on infrastructure the customer owns, so the publisher never receives the data that is analysed. The security page states that there is no vendor cloud holding your rows, no sub-processor to add to a register and no phone-home, and the self-hosting page adds that the interface fetches nothing from a third-party host. Language-model API keys belong to the customer and are stored encrypted with AES-256-GCM. Ownership therefore rests with the customer by architecture rather than by contract: no terms of service, privacy policy or legal notice is published anywhere on the site, so none of this is set out in a written agreement a buyer could rely on.
Reuse rights
Because the platform is installed on the customer's own servers, the publisher has no access to the data and imposes no reuse conditions on it. Queries run against databases the customer already owns, results can be exported freely as CSV, JSON, XML, XLSX, NDJSON or Parquet, and cached or materialised results are written into a database the customer controls. AI calls are the one outbound path: they go to the model provider the customer configures, under that provider's own contract and pricing, and can be kept entirely inside the network by pointing Qrly at a local Ollama or LM Studio endpoint. No published document from the publisher restricts what the customer may do with the data.
Data retention & training
Hosting summary
There is no vendor hosting to describe: Qrly is installed and run on infrastructure the customer owns, whether on-premise, on a virtual machine, in a private cloud or on Kubernetes. Data stays in the customer's own PostgreSQL application database and in the source databases being queried, so the jurisdiction is whichever one the customer already operates in. The publisher frames European data residency as a consequence of that choice rather than as an offer of its own, writing that residency follows because the host is yours. It also states that the software makes no external asset requests, uses no content delivery network for fonts, scripts or styles, and does not phone home, and that there is no sub-processor to add to a register. One distinction matters: the public marketing site getqrly.eu is served from Infomaniak in Geneva, Switzerland. That is the website only and says nothing about where customer data lives, which is entirely determined by where the customer installs the platform.
Things to keep in mind
Risks and trade-offs to weigh before adopting Qrly.
- The company behind Qrly is named on no page of the site; its identity can only be established by matching the published address and phone numbers against the Belgian company register
- No terms of service, privacy policy, legal notice or data processing agreement is published, so no commitment about data, liability or availability is written down anywhere
- The legal entity was registered in 2019 while the product presents itself as established in 2026 — a gap of nearly seven years between the company and the tool
- Self-hosting transfers responsibility for backups, patching, access control and availability to the customer, and the publisher notes that verifying a backup restores is your responsibility
- Sending schema and question context to a cloud language model is a genuine data-exfiltration path; only the local-model option closes it, and that choice is left to the administrator
- Natural-language querying can produce a plausible, well-formatted answer to a subtly wrong question, so the human review step before saving should not become a formality
- The absence of multi-factor authentication and row-level security has to be compensated elsewhere, at the identity provider and in permission design, or it quietly becomes an accepted risk
Setup & Integrations
Technical difficulty
Demanding, and deliberately so. There is no hosted option, so a team must provision a server, run Java 25 and Spring Boot 4 against PostgreSQL 14 or later, and operate the service afterwards. The publisher lowers the first hour considerably: an eight-step wizard probes the database read-only, writes nothing until you confirm, seeds no default account, generates the systemd unit, and never reads a configured secret back. A platform console then covers configuration, logs, memory and scheduled backups. For a team already running Java and PostgreSQL this is routine; for anyone without server operations capability it is out of reach.
Deployment
Integrations
Supported languages
Behind Qrly
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Qrly.
Frequently asked questions
How is the annual fee calculated?
Does the price rise as we add users or embedded viewers?
Is any feature reserved for a higher tier?
Can Qrly run entirely on our own infrastructure?
What does Qrly charge for AI?
Does Qrly support multi-factor authentication?
Is Qrly certified against SOC 2 or ISO 27001?
How many databases can it connect to?
Do the AI agents write to my data?
Which interface languages are available?
Should you pick Qrly?
Qrly is a coherent answer to a real complaint about business intelligence: that dashboards, embedding, alerts, single sign-on and AI arrive as separate products with separate bills. It ships them together, on infrastructure the customer owns, and prices the whole thing on the customer's revenue rather than on seats or queries — so adoption never raises the invoice. The published tranche schedule, from a €180 floor upwards, is unusually transparent, and the refusal to take a margin on AI tokens is consistent with the rest of the positioning.
The engineering ambition is visible: QQL compiling to twelve dialects, four cache tiers ending in change-data-capture, a spreadsheet engine bound to live SQL, and an agent layer where every write waits for a human. The security page is the most persuasive document on the site precisely because it lists what is missing — no multi-factor authentication, no row-level security, no certification whatsoever.
The reservations are real and they are mostly about maturity and paperwork rather than capability. The product dates itself to 2026, has no archived history, names no customer and offers no trial or demo, so a buyer evaluates it on the documentation and a conversation. More seriously, the site publishes no terms of service, no privacy policy and no legal notice, and never names the company behind it — the publisher can only be identified by cross-referencing the postal address and phone numbers it prints against the Belgian company register. For a tool aimed at banks, defence and the public sector, that gap between the rigour of the product documentation and the absence of contractual documentation is the thing to raise first. Qrly deserves a serious look from teams that can run a Java service and want to own their analytics stack outright — with the contracts negotiated directly.
- Choosing a selection results in a full page refresh.
- Opens in a new window.