Notebooks

Real analysis. Not just chat.

An agent that builds a plan, executes it, and documents every decision. Canvas, notebook, and deployment in one workspace.

Agent planningBuilt for data workCanvas + NotebookOutput galleryOne-click deploy

load_sales

load_sales.py

import zerve

df = connect("warehouse/sales_q4").read()

df = df[df["env"] != "test"]

Serverless · 2.1s

Q4 Sales Report

Version 1 · Generated by agent

Describe what you want to build...Build me a Q4 revenue analysis workflow

You direct. The agent executes.

I'll build a Q4 revenue analysis workflow: connect data, prepare it, run a model, generate visualisations, and set up deployment.

Plan

Current Tasks

Connect & prepare data

Build model

Generate report

Deploy workflow

Describe the goal

Plain language in, structured plan out.

Review before anything runs

Auto or manual approval. You choose how much control to keep.

Watch it work

Code, outputs and a full DAG. Every step traceable and rerunnable.

End to end in one place

Notebook, report, deployment. One workflow, full audit trail.

Connect & prepare data

Progress0%

Built for Data Work

Secure connections to data

Connect directly to your warehouse, files or APIs without moving data outside your infrastructure.

Languages, libraries, and frameworks

Python, R, SQL and PySpark. No restrictions on packages, dependencies, or environment.

Scale on demand

Serverless or persistent compute. CPU and GPU instances across Lambda, SageMaker, Kubernetes and ECS.

Files

Snowflake

BigQuery

Notebook

load_sales

clean_segments

join_customers

Not your standard notebook

Each cell is an independent node in the DAG. Its own runtime, cached output, and no shared state. Run anything in any order, or in parallel.

Build model

Progress0%

churn_model

Cells

Generate report

Progress0%

Outputs ready for reports

Every chart and figure your agent produces flows into the gallery — reproducible, linked to its source, and ready to drop into a report.

Every output, automatically captured

Charts, figures and tables flow into the gallery as blocks run. Nothing to export manually.

Reproducible by design

Re-run the notebook and every image updates. Your gallery always reflects the latest results.

One click to a Report

Drop any image straight into a Report from the gallery. The source cell is always linked.

6Image Gallery

Real images created by the agent and members of the Zerve community

Deploy

Progress0%

From notebook to production

Deploy models, APIs, and apps directly from your notebook. Load any output with a single import — no rebuilding pipelines.

app.py

from zerve import variable
 
model   = variable("train_xgboost", "model")
chart   = variable("throughput_analysis", "fig")
df      = variable("load_data", "df")

Load any output:ModelChartDataFrame

churn_model.notebook

Deploy

Framework

Configuration

DNS / Domain

Environment

Production

Compute

Serverless

Preview

churn-model.zerve.app

0.38

Any framework

Streamlit, FastAPI, Dash, Flask. Your code, your stack. Zerve handles the rest.

Live from your notebook

from zerve import variable loads any output directly. No re-running pipelines.

Instant preview

Every deploy gets a .zerve.app URL. Set your own DNS when you're ready.

Ready to start building?

View Pricing

No credit card required · Free to get started