Deepnote research: our notes on building agents

Get started

PLG dashboard live appPM

Weekly signups128k+12%

Activated41%+6%

At risk742-18%

Pipeline by segmentlast 30 days

Enterprise47 accounts

Startup31 accounts

Self-serve92 accounts

published · shared with revenue team

Churn analysisDA

SQL
SELECT variant, churned_30d
FROM ab.assignments a
JOIN events.churn c USING (user_id)

churn rate · 30d

control

4.2%

treatment

7.8%

✦Deepnote AIinsight

Treatment users skipped onboarding 2.4x more often. Segment by first-session length before shipping.

add follow-up cellexplain drivers

top drivers

tutorial_skipped

+0.34

no_team_invite

+0.21

first_session < 2m

+0.14

treatment +85% vs control

Fraud detection agentJJ

✦summarize and cancel fraudulent transactions by country

planning next steps

↓ generated 4 blocks

SQLflag_fraud_txns

SELECT txn_id, country, amount FROM txns WHEREfraud_score > 0.8

PYtrain fraud_detector.pt

ep 010.412

ep 030.187

ep 050.064

DEPLOYfraud_detector → live

▲deepnote.com/api/v1/fraud-detector

postscoring 1.2k txns / seclive

CHARTcard cancellations by country

4 cells · query → train → deploy → visualize

Cursorproject: revenue-app

user

why is March revenue down 12%?

agent

let me check the canonical revenue breakdown skill

⚙ tool_call

deepnote.run(
skill = "revenue_breakdown",
period = "2026-03"
)

project · 4 vCPU · 8 GB

← returned notebook

revenue_breakdown_2026_03.deepnote

4 cells · trace #4729

Visual Studio Codechurn_analysis.deepnote

1import deepnote as dn

3df = dn.query("SELECT …")

4df.groupby('cohort').agg(…)

6fig = px.line(df, x='week', y='retention')

7dn.publish(fig, name="churn-cohorts")

⌘⇧Pcommand palette

>Deepnote: Open in Cloud

>Deepnote: Run as Skill

>Deepnote: Schedule Notebook

✓kernel state14 MB

✓packages47

✓env vars6

✓files3

→ opening at deepnote.com/workspace/…

Claude Code~/revenue-investigation

 ██████╗ ██╗       █████╗  ██╗   ██╗ ██████╗  ███████╗
██╔════╝ ██║      ██╔══██╗ ██║   ██║ ██╔══██╗ ██╔════╝
██║      ██║      ███████║ ██║   ██║ ██║  ██║ █████╗
██║      ██║      ██╔══██║ ██║   ██║ ██║  ██║ ██╔══╝
╚██████╗ ███████╗ ██║  ██║ ╚██████╔╝ ██████╔╝ ███████╗
 ╚═════╝ ╚══════╝ ╚═╝  ╚═╝  ╚═════╝  ╚═════╝  ╚══════╝

 ██████╗  ██████╗  ██████╗  ███████╗
██╔════╝ ██╔═══██╗ ██╔══██╗ ██╔════╝
██║      ██║   ██║ ██║  ██║ █████╗
██║      ██║   ██║ ██║  ██║ ██╔══╝
╚██████╗ ╚██████╔╝ ██████╔╝ ███████╗
 ╚═════╝  ╚═════╝  ╚═════╝  ╚══════╝

user"investigate march revenue dip — return a notebook"

claude

spinning up a deepnote project & pulling the canonical revenue skill

$deepnoteproject create--from-skillrevenue_breakdown

✓ project prj_8f21 · python 3.12 · 4 vCPU · 8 GB

$deepnoterun--cell"join_orders_with_refunds"

✓ 1.4s · 2,184 rows · warehouse: snowflake/prod

$deepnotenotebook publish--shareteam

✓ deepnote.com/n/march-revenue-dip

← handed back

march_revenue_dip.deepnote

7 cells · trace #c12a · ran in 4.3s

The data workspace built for

Everything your team and agents need to explore data, automate insights, and make better decisions.

500,000+400,000+data professionals

AI superpowers for everydata analyst and data scientist

Become a senior data analyst overnight

Describe what you want to do and Deepnote will automagically query, analyze, and interpret data for you. No coding skills necessary.

Deepnote understands your business context and your data stack to deliver high-quality analyses fast.

See in action

Write Python, SQL, R...

Enhance your productivity by using your favorite languages and tools in a single AI-enabled IDE.

...or let Python, SQL, R be written for you

Aware of your code, context, and metadata. Deepnote continuously adapts to turbocharge your work.

Refactor code

Generate code

Explain code

Code completion

Debug & fix

From quick explorations to powerful data apps

Quick explorations

Notebooks

Data apps

Dashboards

ETL pipelines

Data catalog

Knowledge base

Sales report
Customer 360
Churn prediction model
Lead scoring

Sync Salesforce to Notion
A/B test evaluation

AWS cost analysis

NPS analysis

Fine-tune LLaMA 7B
Protein visualization
Forecast sales pipeline

Interactive visualizations

Instantly visualize dataframes as no-code configurable charts.

Beautiful apps & dashboards

Deepnote turns quick data explorations into impressive reports & apps. See some examples.

Arrange into custom layouts

Hide code blocks

Add input blocks and buttons

Schedule notebooks

Schedule notebooks to run hourly, daily, weekly, or monthly.

Deploy notebooks as APIs

Put your notebook into production. Serve your models directly.

Run on powerful GPUs

Choose what kind of hardware you want to use, from CPUs to big GPUs.

Collaborate with the whole team

Deepnote runs seamlessly in the cloud, making environment management and collaboration with your team a non-issue. And sharing work is as easy as sending a link or email invite.

Product manager

Organize your work

Build a library of data projects sorted by folders so teammates can get needed information fast.

Comment, review, version

Have your team comment on blocks to ask questions, provide feedback, and work faster.

Data scientist

Sharing made simple

Share your work with others by simply sending a link or email invite. Or use advanced permission models.

100s of integrations

Connect to your data in seconds with built-in integrations for the most popular warehouses, databases, and lakehouses.

Based on open standards & extensible APIs

See all integrations →

dbt metadata

Run your queries with Jinja in Deepnote’s SQL blocks and browse your company metrics right next to your warehouse schema.

Drag and drop your CSVs

And use them like any other data source.

Spark & Snowpark

Work with terabyte-scale data. Connect to your own cluster or spin up serverless Spark clusters on demand.

Improve your data security with Deepnote

Security gets better and easier with Deepnote. Stop exporting CSVs and taking screenshots of sensitive data. With Deepnote, you have complete control over data access and security.

AICPA SOC, HIPAA, GDPR, CCPA Compliance badges

Secure & compliant

All data is secure and encrypted. Deepnote is compliant with HIPAA and SOC 2, GDPR, and more.

RBAC, SSO, directory sync

Control who can access data and code. Enjoy single sign-on via SAML or OIDC connections and sync your directory.

Custom deployments

Get a secure, compliant, and fully managed Deepnote instance or book a demo to learn about all deployment options.

What makes Deepnote so secure?

Discover how we mitigate risks and protect your data using industry best practices and continuous improvements.

Loved by a community of 500,000+ data professionals

Deepnote gets us from raw SQL to polished visuals instantly, so we can focus on delivering real business value, rather than tweaking charts.

Melvin Low

MLOps platform engineer

Implementing Deepnote was as easy as connecting to BigQuery, and analysts were quickly using Deepnote as their preferred tool for analytics.

Ryan Doherty

Principal Data Analyst

Deepnote has revolutionized our data analysis workflow, enabling our teams to deliver insights in a fast collaborative manner.

Guy Yarkoni

Head of Analytics & BI

Collaborative data analysis where team members can freely share their work and get feedback… This made the analysis workflow much faster.

Khanh Nguyen

Data Analyst

It always surprises stakeholders how fast we work with Deepnote. We discuss something in the morning and we have results to share in the same afternoon.

Thom Hopmans

Senior ML Engineer

I just love how SQL is now a first-class citizen in Deepnote notebooks! 🔥 It is SO easy to query databases!

Charly Wargnier

Developer

Deepnote has become our go-to solution for data science, abstracting away all the infrastructure complexities we previously faced.

Maximilian Strauss

Co‑founder & CTO

We needed a way to instantly dive into data, collaborate seamlessly, and build reusable analytical templates. Deepnote checked all those boxes.

Timothy Chan

Head of Data

At 96 out of top 100 universities

Used by the next generation of data analysts and data scientists.

Learn more ->

Deepnote was incredibly easy to set up and allows us to start new notebooks in seconds. Working together with Deepnote gives us a great window into the ways candidates approach the interview problem.

Becca Carter

Head of Data Science

I just swapped my Jupyter workflow for Deepnote, now open source and fully Apache 2.0. Think of Deepnote as Jupyter, but with SQL, AI, and multiplayer mode, among other handy features.

Lior Alexander

CEO