Connect Supabase. Your dashboard builds itself.

supalens reads your schema, finds the KPIs that matter, and wires up dashboards and alerts on its own. After that — just ask your database.

Start free →How it works →

supalensFileEditViewGoWindowMon, Jul 6 · 10:24

DailyCart — Auto KPI extraction

CONNECT SUPABASE

SupabaseRead your projects · read-only

We read your schema and build a KPI dashboard tailored to your service

Connect SupabaseOAuth · read-onlyConnected

Supabase connected · read-onlyReading schema — 42 tablesAnalyzing relations — orders & checkout

public.orders
public.order_items
public.customers
public.products
public.product_variants
public.carts
public.cart_items
public.payments
public.subscriptions
public.subscription_items
public.coupons
public.refunds
public.reviews
public.shipments
public.categories
auth.users
public.orders
public.order_items
public.customers
public.products
public.product_variants
public.carts
public.cart_items
public.payments
public.subscriptions
public.subscription_items
public.coupons
public.refunds
public.reviews
public.shipments
public.categories
auth.users

42 tablesOrder & checkout schemaSubscription tablesRelations mapped

CORE KPI DASHBOARDLIVE

Daily revenue₩3.1M₩3.8M₩4.7M₩4.2M+12.4%+₩12,400

Revenue · 14 days

Checkout conversion3.8%+0.4%p

Average order value₩38,400−0.8%Alert

Cart abandonment61.2%−2.1%p

Revenue by categoryApparelBeautyFoodLiving

Core KPI dashboard created6 widgets · 2 alerts configured

4Supported databases · Supabase, PostgreSQL, MySQL, Firestore

12Visualization charts the AI picks for you

22MCP tools you use from Claude

100%Read-only · runs SELECT statements only

From connect to auto dashboards, in three steps.

Set up once — the AI writes the SQL and builds your widgets and alerts on its own.

01

Connect your database

Connect Supabase with a single OAuth, or PostgreSQL and MySQL with just credentials — the schema is read automatically.

Connect data source

◗SupabaseConnect with OAuth

42 tables · schema synced

02

Connect MCP to Claude

Add mcp.supalens.ai to Claude Code or Cursor and ask in plain language.

terminal

$ claude mcp add --transport http supalens https://mcp.supalens.ai/mcp

connected · dailycart

03

Auto-generate dashboards and alerts

The AI writes SQL, builds widgets and alerts, and emails you a weekly summary.

Core KPIs◔Alerts 2

Revenue₩ 8.4M

Orders1,204

Trend

Connects right into the AI you already use.

Supalens runs as a single MCP server, mcp.supalens.ai — plug it into Claude Code, Cursor, or Codex and the AI reads your database, queries it, and even builds dashboard widgets.

Setup guide →

Claude · supalens MCP

Compare this week's cart abandonment rate with last week.

⏵execute_sqlsupalens

SELECT date_trunc('week', c.created_at) AS week,
       1 - count(o.id)::float / count(c.id) AS bounce
FROM carts c
LEFT JOIN orders o ON o.cart_id = c.id
WHERE c.created_at >= now() - interval '14 days'
GROUP BY 1 ORDER BY 1;

Last week63.3%

This week61.2%−2.1%p

This week's abandonment rate is 61.2% — a 2.1%p improvement over last week.

⏵add_widgetsupalens

Saved it as a widget on your core KPI dashboard.

One question, all the way through.

From writing SQL to visualization, dashboards, and alerts — Supalens links the start and end of your data work into a single flow.

01

Natural language → SQL

Claude reads your schema and code context (SUPALENS.md) to write grounded, well-reasoned SQL.

Show me last week's sales by category for DailyCart

⏵ View SQLselect category, sum(amount) …

02

Multi-DB routing

Automatically routes across Supabase, PostgreSQL, MySQL, and Firestore.

SupabasePostgreSQLMySQLFirestore

Demo · E-commerce sample▾

03

Automatic dashboards

Turns query results into any of 12 chart widgets and gathers them on a board.

Daily revenue

Alerts + root-cause analysis

Watches thresholds, change rates, and anomalies — when one fires, AI analyzes the cause and notifies you by email and Slack.

🔴Daily revenue−26%

Analyzing cause…

05

Weekly digest

Every Monday morning, AI summarizes the key changes versus last week and emails them to you.

supalens · Weekly digest

Sales last week +12%
Churn improved
New signups slowing

06

GitHub context

Reads your repo's code and schema to build context and generate more accurate SQL.

SUPALENS.mdGenerated from repo

# ordersCompleted payment = status = 'paid' onlyExclude refunds via refunded_at

Alerts · AI root-cause

It watches your metrics — and when one breaks, it sends the cause too.

Define a SQL condition and Supalens checks the value on every schedule. The moment it crosses your threshold, AI points to the related deploys and logs, adds a likely-cause hypothesis, and notifies you over Slack and email. When it recovers, the all-clear follows automatically.

Slack · #alerts

supalensAPP2:14 PM

Daily revenue below threshold

₩3.1MBaseline ₩4.2M−26%

Metric: daily_revenue
Condition: < ₩4.0M
Source: DailyCart · prod

Root-cause analysis

Since the coupon migration deployed on 7/5 (0051_coupons.sql), the payment failure rate has risen over the same window.

Checkout webhook retry logs spiked at the same time — we recommend reviewing the payment-confirmation step.

When the anomaly recovers, an all-clear alert follows automatically.

A few questions became this board.

Ask a question and Supalens picks the right chart from 12 types and draws it into a dashboard for you.

Explore dashboards →

DailyCart · Core KPIs

Today's revenue

vs. yesterday

Today's revenue

₩4.2M

12.4%vs previous

Today's orders

vs. yesterday

Today's orders

1,284orders

8%vs previous

Daily revenue trend · 14 days

Product	Revenue	Orders
Organic apples 1.5kg	₩4.2M	612
Jeju Samdasoo 2L×6	₩3.2M	540
Hanwoo sirloin 300g	₩2.9M	188
Avocado 4-pack	₩1.8M	470
Antibiotic-free eggs 30ct	₩1.3M	388

By design, Supalens can't change your data.

Read-only isn't a promise — it's the architecture. There's no tool that writes, your credentials are sealed, and the analysis AI runs isolated, without them.

ENCRYPTION

Sealed credentials

Your database credentials are envelope-encrypted with a per-organization key (AES-256-GCM). The master key that wraps it is never stored in the database, so a database dump alone can't decrypt them. Plaintext never reaches your browser or the AI model.

AES-256-GCMper-tenant DEKAAD-bound

ACCESS

Structurally read-only

Every question runs read-only. Non-SELECT statements are rejected, and a READ ONLY transaction, statement timeout, and row cap are always applied. For directly connected databases we mint a dedicated read-only role with least privilege.

SELECT-onlyREAD ONLY txrow cap

RUNTIME

Isolated AI execution

The analysis AI runs in a sandbox with credentials stripped from its environment — no shell, file, or network tools. Database credentials never enter the model process, and only pre-approved tools can run.

env-strippedno shell / netdeny-by-default

ISOLATION

Tenant isolation

Row-level security is enabled on every user table. One analysis is bound to one organization, chosen by the server — never by the question or the model, so it can't reach another org's data.

RLS1 run = 1 tenant

In the open

We don't replicate your whole database. During analysis, query results are sent to Anthropic (Claude), and the answer plus a sample of up to 200 rows and your schema structure are stored. Every query is audit-logged and auto-deleted after a set period, and deleting your account destroys the keys — making encrypted credentials unrecoverable.

Compliance

SOC 2In progress