Talk to your data.
Get answers in seconds.
For every team that works with data. Your data, always under your control.
Ask in plain language — the assistant queries your data, builds charts, and prepares reports. Your data stays on your servers, every query logged.
Geçen ay en çok satan 5 ürünü göster
Salt okuma
Veriye dokunulmadı
Sorgu kaydedildi
Kim, ne, ne zaman
Veri ağdan çıkmadı
Yerel model aktif
Your data stays yours · Every query is logged · Runs on your servers
For every team. You stay in control.
0%
Service availability
0+
Database engines supported
0-bit
Bank-grade encryption
0+
AI models — local supported
Compliance
Veri tabanları
Ecosystem
Demo Preview
Full demo video coming soon
In the meantime, watch a live preview of how nivq works — or join a one-on-one demo with our team.
Show me the top 5 best-selling products last month
Salt okuma
Veriye dokunulmadı
Sorgu kaydedildi
Kim, ne, ne zaman
Veri ağdan çıkmadı
Yerel model aktif
Tam demoyu sizinle birlikte yapalım
Beta erken katılımcıları için kişisel walkthrough — kendi verinizi getirin, kendi sorunuzu sorun.
Getting Started
Up and running in minutes
01
Connect Your Database
Add your database credentials in seconds. PostgreSQL, MySQL, SQL Server, and more — all encrypted end-to-end.
02
Select Your Tables
nivq automatically discovers your entire schema. Simply choose which tables the agent can access.
03
Start Asking Questions
Type your question in plain language. nivq generates accurate queries, runs them, and returns results — instantly.
Insight workflow
From a question to a shared insight
The motion the demo above shows, broken into four named beats. This is what your team will actually do every day in nivq — no SQL, no copy-paste, no waiting on the data team.
Ask
Type a question in plain language. Turkish, English, mixed — nivq picks the right tables, joins them, and runs the query.
Last quarter's top 10 customers by revenue
Chart
Need a visual? Just say so. Bar, line, pie, area — auto-picked from your data shape, themed to your workspace, rendered live in the chat.
Pin
Save the answer to a dashboard. Drag to reorder, resize freely, group into multiple boards. KPIs that stick around even after the chat is archived.
Share
One link to the live chart on a polished public page. No login for the recipient, no PDF attachment, no stale snapshot. Revoke anytime.
Two Approaches
AI assistants work in two different ways
The same term — "AI agent" — is used for very different jobs. Knowing the difference helps you pick the right product. Here's where nivq stands and why.
Marketing automation, sales bots, social media managers
The assistant decides for you and acts on your behalf.
- Pauses ad campaigns, changes budgets, updates customer records, posts to Slack — around the clock, without your approval each time.
- Fast and effective for marketing and customer operations. But when something goes wrong — like the April 2026 incident where an AI agent deleted a production database — recovery can be costly.
- Generally cloud-based, optimized for fast action. Not suitable when every step needs to be auditable.
nivq
The assistant works for you, but you stay in control.
- The assistant understands your question, reads the data, writes the query, builds the chart — but it cannot touch your database without your approval, and never takes any action on its own.
- If you ever grant data modification capability, every change goes through an approval screen: what will change, how many rows are affected, can be reverted with one click.
- Runs on your own servers. Can work entirely offline. Every query, every approval, who did what — all logged.
Autonomous assistants are powerful tools for marketing and customer operations. Controlled assistants are for every team where data matters — from a small startup to the largest enterprise. nivq picks this second path — without giving up speed, without taking control away from you.
What You Get
Everything you need for safe AI on your data
Accurate answers, your data always under your control, and full visibility into who asked what. For every team that works with data — built to the strictest security standards.
Reads, Never Writes
The assistant cannot delete, update, or insert anything in your database. Three layers of protection block any modification attempt — even by mistake. If you ever want write capability, every change goes through your approval first.
Every Query Logged
Who asked, when, what question, what answer — all stored. Export to Excel for your auditor with one click. Default retention 7 years, fully configurable.
Choose Your AI Model
Use OpenAI, Claude, Gemini, or Mistral — or skip the cloud entirely and run a local model on your own server. If you want zero data leaving your network, this mode is for you.
Complete Isolation
Even when multiple companies share the same system, no one can see anyone else's data. Five independent security layers guarantee this; each is verified by automated tests.
Learns With Your Approval
The assistant learns from successful queries and gets better over time — but every learning step requires admin approval. A wrong answer cannot be silently memorized as 'right'.
See How It Thinks
Watch the assistant understand your data, decide which tables to look at, write the query — all in real time, in front of you. No black box.
Why nivq
Three things that set nivq apart
Not just a chatbot, not just a model interface, not an agent that takes action behind your back. nivq learns your data, writes its own queries, runs them — and always stays within the boundaries you set.
Gets smarter with every query
nivq doesn't just translate your question — it remembers. Every interaction teaches the agent your specific data model, business terminology, and query patterns.
- Learns column relationships, naming conventions, and domain-specific terms automatically
- Accuracy improves week over week without any manual training or feedback
- Builds a knowledge base unique to your organization's data
Your AI, your rules
Choose any LLM provider — or switch anytime. nivq gives you full control over which model powers your queries, with your own API keys and zero vendor lock-in.
OpenAI
GPT-4o
Anthropic
Claude
Gemini
Azure
OpenAI
Mistral
Large
Ollama
Local
+ Any OpenAI-compatible endpoint
- OpenAI, Anthropic, Google Gemini, Azure, Mistral, DeepSeek — all supported
- Run fully local with Ollama for maximum privacy and zero cloud dependency
- Switch providers in seconds — no migration, no reconfiguration
Watch it think in real time
nivq doesn't just return results — it shows you the entire reasoning pipeline live. From understanding your intent to generating and executing queries, every step is transparent.
- Live pipeline view: intent analysis → schema mapping → query generation → execution
- See the generated query before it runs — full transparency, no black box
- Results stream in as they arrive — no waiting for the full query to complete
Security & Compliance
Your data stays yours, the audit trail is always ready
Built for banks under regulatory scrutiny, insurers under data protection laws, multinationals exposed to AI regulations. Security is not a marketing claim — it's the architecture itself.
- Your data never leaves the network — runs on your servers with local AI model support
- Every query, every answer, every denied access is stored — auditors get an Excel report in one click
- Bank-grade encryption — both for storage and transmission
- Different companies on the same system are fully isolated — five independent layers, all tested
- Annual licence model, can be hardware-bound, time-rollback attempts detected
On Your Servers
Works offline-capable
Full Audit Log
7-year retention, exportable
Enterprise Login
Keycloak, Okta, Auth0
Full Isolation
Five-layer separation
Banking Mode
Your bank's policy says "no vendor connects directly to the database"? This mode is for you.
Some institutions have strict security policies: external vendors can never connect directly to the database. In that case, nivq routes the query through a proxy server you control. We never communicate with your database directly; you get a corporate gateway where you can place your own audit and policy layer.
nivq
Yapay zeka asistanı
Aracı Sunucunuz
Sizin denetiminiz
Veri Tabanınız
Kendi ağınızda
No direct connection
When this mode is on, every query the assistant writes is sent to your own server as a request. nivq holds no database credentials whatsoever in this mode.
Your own audit layer
Row-level filtering, query rewriting, sensitivity tagging — all done inside your proxy server. nivq simply receives the result and shows it to you.
Fully closed system
Combine this mode with a local AI model (Ollama, Mistral) and your questions, your schema, and your results never leave your network.
Two connection methods
Long-lived event stream or one-shot request — works with whichever your system speaks. Both standards supported.
Proxy server contract
tool: execute_query // sorguyu çalıştır, satırları döndür
input → { query: string, namespace?: string }
output → { success: bool, rows: array, rowCount: int, error?: string }
tool: get_schema // veri tabanı türü, tablolar ve sütunlar
input → {}
output → { dialect: string, namespaces: [{name, tables: [{name, columns}]}] }
bağlantı: olay akışı · tek seferlik istekComparison
What's different from existing tools
Traditional dashboards and generic chatbots have the same problem: the person asking the question has to learn the tool, and the tool doesn't know your data. nivq fills the gap between them.
Traditional dashboards
Tableau, Looker, Power BI
- Every new question needs a data analyst or someone who knows SQL.
- Dashboards age — fields get renamed, relationships shift, charts silently show wrong numbers.
- Per-seat licensing is expensive for managers who only check a couple of reports per month.
Generic chatbots
ChatGPT, Claude, Gemini
- Don't connect to your database — you have to copy-paste rows and hope the model doesn't make up an answer.
- Don't remember your tables, your relationships, or your business terms — you have to explain everything from scratch every time.
- Sharing a result means taking a screenshot — no live link, no idea who saw it, can't be revoked.
nivq
AI assistant that knows and learns your data
- Ask in plain language, get a real query that runs on your live database — no copy-paste.
- Learns and remembers your tables, relationships, and business terms — no need to re-explain.
- Pin a chart to a board, share a live link, revoke anytime — recipients don't even need to log in.
Use Cases
Built for everyone who needs data
From business analysts to engineering leads — nivq gives every role instant access to database insights without writing queries.
Ad-Hoc Analysis
Business Analysts & Product Managers
Run complex queries without waiting for the data team. Ask revenue breakdowns, funnel metrics, or cohort analysis in plain language.
"Show me monthly revenue by region for the last 6 months, with growth percentages"
Accelerated Data Exploration
Data Engineers & Data Scientists
Skip the boilerplate queries for exploratory analysis. Use nivq for quick data profiling, schema exploration, and hypothesis testing.
"What's the distribution of order values for customers who signed up in Q1?"
Instant KPI Monitoring
C-Level & Department Heads
Get real-time answers to business questions without dashboards or reports. Ask about performance, trends, and anomalies directly.
"How does this month's churn compare to the same period last year?"
Database Health Checks
DevOps & Platform Engineers
Query operational metrics, check table sizes, monitor slow queries, and investigate production issues — all in natural language.
"Which tables grew more than 20% in the last 30 days?"
Developer Access
Embed it in your own app — or call it through the API
nivq is more than a web UI. Open a chat door to your end users from inside your own product, or build a custom integration through the API. Your data stays on your servers; the assistant runs under your brand.
End-user chat inside your own product
Drop a single script tag into your page and nivq chat lives inside your app. Your brand colors, your logo, your isolation rules. End users ask in plain language without ever leaving your product.
<script src="https://cdn.nivq.ai/embed.js" data-key="nvq_pk_…" data-workspace="acme"> </script>
- Brand colors, logo, tone — fully yours
- Per-user isolation (each end user sees only their data)
- iframe or web component — your choice of frame
- Ready-made SDK: React, Vue, vanilla JS
Programmatic access via REST + webhooks
Chat, query, chart, audit — all available via API. Call from your backend, feed your own dashboard product, wire it into automations. Your API key, your business.
curl https://api.nivq.ai/v1/chat \
-H "Authorization: Bearer $NIVQ_KEY" \
-d '{
"agent": "sales_db",
"question": "Bu çeyrek geliri göster"
}'- REST API — chat, queries, audit logs
- Webhooks — fired when a new answer is ready
- Official SDKs: Python, Node.js, Java, Go
- Anthropic-compatible prompt caching — token savings out of the box
Integrations
Works with your stack
Connect to the databases and AI providers you already use. No adapters to install, no plugins to configure.
Supported Databases
PostgreSQL
MySQL
MariaDB
SQL Server
Oracle
SQLite
and more
LLM Providers
OpenAI
Anthropic
Google Gemini
Azure OpenAI
Mistral
DeepSeek
Ollama
Custom API
and more
Pricing
Simple, transparent pricing
Start free and scale as your team grows. No hidden fees, no per-query charges.
Free
Free forever. Explore NivQ with a minimal quota and a curated entry-level model.
Solo
For individuals and freelancers running focused analytics.
Team
Everything your growing team needs — priority models, more seats, faster support.
Business
For data teams that need higher throughput and premium models.
Enterprise
Dedicated SLA, SSO, BYOK flagship models, on-prem and private deployment.
No credit card required to start. Prices are subject to change — existing subscribers keep their current rate.
Compare all plans
| Feature | Free | Solo | Team | Business | Enterprise |
|---|---|---|---|---|---|
| Queries | 15/day | 50/day | 250/day | 750/day | NaN/day |
| Database Connections | 1 | 2 | 5 | 15 | NaN |
| Users | 1 | 1 | 5 | 15 | NaN |
| Workspaces | 1 | 1 | 2 | 5 | NaN |
| Automatic Table Discovery | |||||
| OpenAI Support | |||||
| Ollama (Local AI Model) | |||||
| All AI Providers (Anthropic, Google, Mistral, DeepSeek) | |||||
| Multiple Workspaces | |||||
| Role-Based Access Control | |||||
| Self-Learning Assistant (Query Patterns & Business Rules) | |||||
| Priority Support | |||||
| Enterprise SSO (SAML/OIDC) | |||||
| Advanced Audit Logs | |||||
| On-Premise Deployment | |||||
| Dedicated SLA |
FAQ
Common questions
Can the assistant change anything in my database?
No. By design, the assistant is read-only — it cannot delete, update, or insert. Three layers of security checks block any modification attempt. If you ever enable write capability in the future, every change goes through your approval first: it shows you what will change, how many rows are affected, and you can roll it back with one click.
Does my data leave my network?
It depends on your choice. If you use cloud AI models (OpenAI, Claude, etc.), your questions are sent to that model. If you have sensitive data — banking records, insurance policies, personal information — you can switch to a local model (Ollama) running on your own servers. In that case, no data leaves your network at all.
Is it compatible with KVKK, BDDK, and the EU AI Act?
Yes — the technical controls these regulations require are in place: option to run on your own servers, all queries logged (default 7 years), full data isolation between companies, personal data minimization. Independent certifications (SOC 2, ISO 27001) are scheduled. Detailed mapping is available in our compliance documentation.
How much setup is required?
Very little. Enter your database connection info, pick which tables the assistant can see, and you're started. The assistant scans your data structure itself and learns your terminology as your team uses it. Or click "Try with sample data" on first launch and chat with a pre-configured assistant immediately.
Which databases are supported?
PostgreSQL, MySQL, MariaDB, SQL Server, Oracle, SQLite, CockroachDB, Amazon Redshift, Snowflake. If your bank's policy forbids any vendor from connecting directly to the database, the assistant can route queries through your own proxy server — in this mode, nivq never communicates with your database directly.
Can I use my own AI model?
Yes. OpenAI GPT, Anthropic Claude, Google Gemini, Azure OpenAI, Mistral, DeepSeek, or your own Ollama installation — all supported. On Enterprise tier you bring your own API key, costs are billed to your account. Anthropic's smart caching is on by default, so a typical query costs 85–90% less for the cached portion.
Does the assistant get better over time?
Yes — but in a controlled way. Every successful query becomes a candidate for the assistant's 'learning list'. After admin approval, it gets memorized permanently. This way, a wrong answer cannot be silently taught as 'correct' without your knowledge. Issues seen in other text-to-SQL products (like the 'poisoned memory' problem in Vanna AI before it was archived in 2026) do not happen here.
What happens if the licence isn't renewed?
It doesn't fully shut down. Thirty days before expiry, warnings start, with daily admin emails. On the expiry date, the system enters read-only mode — existing data can be viewed, no new operations. After 14 grace days, only licence renewal and login screens remain accessible; data is never deleted, audit logs are preserved, and the assistant's learnings are kept. The moment you install a renewed licence, everything is back to normal in seconds.
Ready to meet nivq?
Create an account, click "Try with sample data", see the assistant in five minutes. Pilot on your own data when you're ready.