Thor

From question to dashboard in one flow.

Thor turns business questions into trusted answers, generated reports, reusable insights, and dashboards while staying compatible with multiple databases and enterprise security requirements.

Ask in plain English

Generate report + SQL + data

Build dashboards instantly

Customize with skills

Q&A

Natural language in, structured work out.

Analysts and operators can start with a question instead of a ticket or a BI backlog.

Dashboards

Insight can graduate into a dashboard.

Saved findings become reusable building blocks instead of disappearing in chat history.

0 panels

Trust

Every answer can show the proof.

Thor can package the narrative report, the SQL, and the underlying data behind the conclusion.

3 output modes

Rules

Teams can teach Thor their rules.

Business logic, metric definitions, filters, and table knowledge all stay reusable.

Supported Now

Connect across the database systems your team already runs.

Version support stays explicit, compatibility feels productized, and upcoming non-relational support is visible without shouting.

PostgreSQL

Version 12 and above

12+

MySQL

Version 8.0 and above

8.0+

SQL Server

SQL Server 2019 and above

2019+

Oracle

Oracle 19c and above

19c+

Aurora

PostgreSQL and MySQL compatible

PG / MY

More coming

NoSQL and non-relational engines are next on the roadmap

Enterprise-grade data protection is built into how Thor connects.

Your data does not need to leave your infrastructure to become useful. Thor connects read-only, processes queries in real time, and is designed for teams that need safety, privacy, and operational control from day one.

Read-only

Encrypted transport

No data retention

Read-Only Access

Thor only needs SELECT permissions. No writes, no schema changes, and no operational risk to production data.

End-to-End Encryption

Connections can use TLS in transit, and credentials can stay encrypted at rest so operational setup does not rely on plaintext secrets.

Zero Data Retention

Query results are streamed for the answer in front of the user and do not need to become a copied warehouse of sensitive customer data.

Atlas Ops DB · PostgreSQL 15.2

Yang

Add to Quick Start

Thor AI

Progress

Analyzing schema0.3s

Selecting tables: orders, payments0.3s

Generating optimized SQL with self-correction0.8s

Building report1.2s

Analysis Details

Identified the orders table and joined payments to confirm recognized revenue.

Filtered to delivered orders to exclude cancelled or pending transactions.

Flagged November as an anomaly against the rolling growth trend.

Key Findings

Total 12-month revenue reached $16.08M with a strong growth arc through October.

November dropped 30.6% month over month while order volume stayed stable.

Recovery resumed in December, but results remained below the October peak.

Full Report

SQL

Raw Data

Saved · Used in Dashboard

Monthly Revenue from Delivered Orders (Past 12 Months)

Total revenue across the past 12 complete months was $16,076,314.34, with a clear growth phase in the first half followed by a notable dip and partial recovery.

Key Insight: June to July accelerated after the Pro tier pricing adjustment, while November behaved like a pricing or refund anomaly rather than a demand decline.

Highlights

Revenue climbed from $981K in March to $1.69M in October.

The strongest growth period was June through August, averaging +14.2% month over month.

November revenue fell by $516K with normal order volume, suggesting a billing-side issue.

Safety And Control

Thor is designed to keep SQL generation safe, reviewable, and under control.

Thor uses an internal self-check loop to help ensure generated SQL is usable and accurate. On top of that, SQL and raw data stay visible so teams can perform a final quality check and keep human judgment in control of the answer.

Evidence mode

Share-ready

Audit-friendly

Designed for trust, not just speed.

Teams can start in plain English and still land on something reviewable. That makes Thor useful both for quick exploration and for work that needs to stand up in front of stakeholders.

Business-ready summaryClear narrative output helps teams understand the result before sharing it more broadly.

Inspectable SQLThor runs a self-check process on generated SQL so the query path is more reliable before it reaches the user.

Human quality checkSQL and raw data remain visible so users can verify the output and keep final answer quality under human control.

Report

SQL

Data

A reviewable answer, with evidence attached.

Thor returns a readable business answer, but it also keeps the supporting materials visible so teams can verify quality before they act on the recommendation.

Why this matters: The answer can be checked by both business users and technical reviewers before it is shared or turned into a dashboard.

How quality stays controlled

Thor presents the business conclusion in plain language first.

The generated SQL remains available for technical review and validation.

Raw data stays accessible so a human can perform a final quality check.

Built for reliable execution: Thor includes a critical execution and self-correction loop to help ensure generated SQL is usable and accurate before results are returned.

SELECT DATE_TRUNC('month', o.delivered_at) AS month,
  SUM(o.total_amount) AS revenue
FROM orders o
JOIN payments p ON p.order_id = o.id
WHERE o.status = 'delivered'
  AND o.delivered_at >= NOW() - INTERVAL '12 months'
GROUP BY month
ORDER BY month ASC;

month	revenue	mom_change
2025-03-01	981327.63	NULL
2025-06-01	1338578.76	16.44
2025-10-01	1685773.46	1.58
2025-11-01	1169662.30	-30.62
2025-12-01	1358088.30	16.11
2026-02-01	1241386.24	-0.32

12 of 12 rowsAtlas Ops DB · PostgreSQL

Thor

Generating dashboard

0%

Building

Crafting your dashboard

Analyzing schema, arranging layout, and packaging insights for the team.

Schema

Layout

Charts

Insights

Layout assembled

Thor arranges saved insights into a dashboard structure that is ready to scan.

Charts selected

The system decides where trends, comparisons, and anomalies are best communicated visually.

Highlights packaged

KPI tiles, chart annotations, and anomaly callouts are prepared for the final reveal.

Revenue

$16.1M

+26.5% YoY

Anomaly

-30.6%

Nov dip

Generated Dashboard

Thor turns the most important questions into a dashboard your team can use every day.

Thor automatically selects the most important questions, metrics, trends, and anomalies to display. Customers can also save any valuable answer and place it on the dashboard, making the issues they care about most easy to understand at a glance.

6 metrics

2 charts

1 anomaly callout

Total Token Usage

$0

-90.4% vs last month

Total Conversations

0

+12.3% this month

Registered Users

0

+8.1% total users

Active Tenants

0

+5.2% organizations

Monthly revenue - past 12 months

Growth through October, anomaly in November, recovery after the dip

Generated from saved insight

November anomaly: revenue dropped by 30.6% while order volume stayed stable, pointing to pricing, refund, or billing-side behavior rather than demand contraction.

Token spend and usage

Input and output token mix by month

May

Jun

Jul

Aug

Peak month$1.69M

Average$1.34M

Nov anomaly-30.62%

Recovery$1.36M

Revenue highlights

$0

Total revenue over the last 12 months

Peak month reached October 2025 at $1.69M.

Growth phase from March through October totaled +71.8%.

Recovery resumed in December but remained below the pre-dip peak.

Shareable outputs

The same dashboard can still link back to the full report, the SQL, and the underlying dataset for review, while staying tailored to the questions each customer cares about most.

Thor automatically promotes the most important questions and metrics into dashboard-ready views.

Customers can save valuable answers from chat and customize what appears on the dashboard.

Teams can export, refresh, and iterate without rebuilding work.

Skill Customization

Thor can learn the rules that matter to your business.

Instead of asking users to remember every metric definition, teams can encode recurring logic, filters, and domain rules as reusable skills that shape future answers.

Reusable rules

Multi-database scope

AI optimize

All User System

learnedHigh

Revenue = SUM of paid orders only

Yang51x used

patternMust

Soft Delete Filter on All Tables

System112x used

calcMedium

MRR from subscription amounts

Yang22x used

relationHigh

Order to Payment Link via stripe_id

System19x used

Thor in one sentence

Ask. Validate. Generate. Reuse. Customize.

Thor helps teams move from natural-language questions to evidence-backed answers, shareable reports, generated dashboards, reusable business logic, secure database connections, and domain-aware analysis without breaking the operational flow they already trust.

Start

Plain-English Q&A

Operators can begin with a business question instead of waiting for a dashboard request.

1 prompt

Prove

Report, SQL, and data together

The answer can be shared with confidence because the supporting evidence is visible.

3 output views

Scale

Generated dashboards across connected sources

Useful answers can be promoted into persistent dashboard surfaces while Thor stays aware of the databases behind them.

Multi-source ready

Adapt

Custom skills, security, and domain rules

Thor can be taught the data, access rules, and metric definitions that make your environment unique.

Version-aware