Hex vs Deepnote: a side-by-side comparison

Hex and Deepnote both combine code, data, AI, and outputs in a browser-based workspace, but they’re built for different kinds of work.

Hex is a BI and analytics platform focused on exploring warehouse data and turning results into dashboards, reports, and interactive apps. It combines a notebook for SQL and Python or R with an app layer for presenting outputs, and includes tools like Threads, Explore, and semantic models for self-service analysis. Hex runs in Hex: there is no local execution, and projects don't run anywhere else once they leave.

Deepnote is a full data workspace for teams that go beyond BI into data science and engineering. It provides a powerful, cloud-based workspace that allows users to easily explore, collaborate on, and share data, create interactive charts and dashboards, and build and deploy machine learning models. Deepnote Agent works inside the project, and Deepnote MCP opens the same projects, notebooks, blocks, runs, and integrations to external agents such as Codex and Claude. Projects also open and run in VS Code, Cursor, etc., and convert to and from Jupyter, Quarto, and marimo.

The key difference: Hex is centered on BI, self-service analysis, and data applications. Deepnote gives technical data teams a broader workspace for data analysis, data science, machine learning, scheduled workflows, and local or cloud development.

Hex vs Deepnote at a glance

What is Deepnote?

Deepnote is a collaborative data notebook platform organized around projects. A project can contain several notebooks that share files, data integrations, packages, permissions, and compute.

That structure lets a team divide one piece of work into related parts. A customer-retention project might contain separate notebooks for exploratory analysis, feature engineering, model training, inference, and a stakeholder-facing application. The notebooks use the same project resources without forcing the entire workflow into one document.

As we describe in the notebook manifesto, a notebook should preserve the code, outputs, context, and execution record behind a result. Deepnote’s project model keeps those parts together while allowing a larger workflow to be divided across several notebooks.

A project can later support several delivery paths:

• A data app presents selected inputs and outputs while keeping the implementation in the notebook.
• A scheduled notebook or API-triggered run turns the analysis into recurring or application-driven work.

Both remain backed by the notebook the team continues to edit, so a change to the analysis does not need to be reproduced in a separate reporting layer.

Deepnote also supports local development through its official extensions for VS Code-compatible editors and projects can be converted to and from Jupyter, Quarto, and marimo, which keeps the work reviewable outside the browser. Teams can edit and execute notebooks locally, then use Deepnote Cloud for shared compute, collaboration, Agent, schedules, and apps.

What is Hex?

Hex is a data analytics and BI platform built around projects that pair a notebook with an App builder. Each project combines a Notebook view, where analysts work with SQL and Python or R, with an App builder for turning selected results into reports, dashboards, and interactive applications.

The Notebook view supports SQL, Python, R, charts, pivots, input parameters, writeback cells, and UI-driven transformations.

Hex also has workspace-level assets around those projects:

• Components package reusable groups of cells for use across projects.
• Threads give business users a conversational way to query warehouse tables, semantic models, and endorsed projects.

Hex Apps provide a drag-and-drop presentation layer with tabs, inputs, drill-down, notifications, and different access levels for editors and consumers.

Hex is a closed-source platform, which creates more vendor lock-in around the full project experience. Projects can be exported, but some Hex-specific functionality does not carry over cleanly to Jupyter, and the full Notebook and App environment remains hosted in Hex.

Deepnote vs Hex architecture

Deepnote treats a project as a container for related notebooks. Those notebooks share the same integrations, filesystem, dependencies, machine settings, and access controls. Deepnote Agent works in that same environment with the same project context. Deepnote execution modes let a user run one block or run that block with everything downstream that depends on it. Deepnote infers those relationships through a DAG and provides a dependency-graph view.

This works well for analyses that have several stages. The data-preparation notebook can remain separate from model training, while an inference notebook runs on a schedule and another notebook provides the app. Project version history covers the notebooks together, which matters when they depend on each other.

A Hex project centers on one Notebook view and its published App. Teams can divide a long notebook into sections or move reusable cell groups into Components. New Component versions do not automatically replace the versions already imported into downstream projects; users review and apply the update.

Both approaches can support complex work, but they encourage different project shapes. Deepnote gives a team more room to divide a larger workflow across notebooks without splitting its data, environment, and permissions across separate projects. Hex gives one analysis a tightly integrated notebook-and-app surface.

Deepnote Agent vs Hex AI

Deepnote Agent

Deepnote Agent opens as a panel in the project and works across the whole notebook rather than one cell at a time. It has two explicit modes.

• Ask mode answers questions without changing the notebook. It can explain code, discuss an approach, inspect the available data, or help with Deepnote features.
• Edit mode can add, change, move, or remove blocks. It executes code, examines the output, and adjusts the analysis when the result requires another step.

Agent creates a visible plan before carrying out larger requests. Its action list shows which blocks it added, edited, or executed. When the work finishes, the user can review the changes and undo the complete run.

This is useful when the notebook contains shared or scheduled logic. A teammate can discuss a change in Ask mode, switch to Edit when the plan is clear, and review what Agent changed before keeping it. Teams can pick the model behind Agent or leave it on Automatic, and Agent can use the project’s connected MCP integrations as well.

Hex Notebook Agent

The Hex Notebook Agent also works with the project’s code, outputs, data connections, semantic models, and workspace rules.

It splits the job across several agents aimed at different people.

• The Notebook Agent generates and edits Python, SQL, Markdown, pivot, input, and chart cells with the project’s structure and warehouse schema as context, and can take other Hex projects as additional context by mention or URL.
• Threads answers questions conversationally for business users, prioritizing endorsed and semantically modeled data.
• Chat with App lets a published app’s audience summarize it, change inputs, and re-run.
• The Modeling Agent builds semantic models for admins and managers, and Generative Apps, in beta, builds an app from a description.

The agent can move and delete cells, debug code, search the web, and arrange or publish the project’s App. Model and effort controls let users select the underlying model for each thread.

Deepnote vs Hex: MCP and external agents

Deepnote and Hex both connect to external AI clients, but the operations exposed through MCP differ.

Deepnote MCP exposes the workspace itself and lets Codex, Claude, Cursor, VS Code, and other compatible clients search projects, create projects and notebooks, create and update blocks, reorder them, start runs, read run snapshots, list historical runs, inspect integrations and their table structure, trace which notebooks and blocks use a given integration, and search Deepnote’s documentation. The agent works under the permissions of the authenticated user or API key.

This means an external agent can create a notebook, execute it, inspect the output, and write the result back into the same workspace. Our walkthrough of Codex for data analytics with Deepnote shows that workflow across existing projects, models, scheduled analyses, and apps.

Hex splits the same territory between two tools. The Hex MCP server focuses on searching projects and creating, retrieving, or continuing Threads. That covers asking questions from Claude, Cursor, ChatGPT, Codex, Glean, or Figma, etc.

Hex’s CLI, which manages projects, cells, runs, connections, threads, and context, and which a local agent can drive, provides the separate path for creating and modifying projects and cells, including through its Codex integration.

For teams that want an external coding agent to operate directly on notebook blocks and runs, Deepnote exposes the shorter route.

Where Deepnote fits best

Deepnote fits teams whose work spans exploration, modeling, and delivery. Analysis that changes direction as results appear, hybrid SQL-into-dataframe-into-chart workflows, feature engineering and model training on hardware sized for the job, and serving from the same project all sit in one environment, without a separate tool for the parts that need a real machine.

Agent and MCP operate inside that same project structure. People and external agents use the same integrations, permissions, compute, files, and run history rather than recreating the analysis in a separate agent environment.

Deepnote is a good fit for:

• data science or machine-learning workflows that move from exploration to training, inference, scheduling, APIs, and apps
• direct notebook access from Codex, Claude, Cursor, or another MCP client
• teams that want local IDE development alongside managed cloud collaboration

The artifact a data scientist builds is the artifact a stakeholder opens, so there is no second copy of the logic to keep in sync.

Mixed-language teams are the other clear case. R, Python, and SQL coexist in a single Deepnote notebook, which matters for groups where statisticians and engineers work on the same analysis.

Where Hex fits best

Hex is better aligned with BI and self-service analytics over warehouse data.

Analysts can build SQL-backed reports and interactive Apps, while Threads, Explore, and Chat with App let consumers query or drill into data without touching the project's logic, and teams that already model metrics in dbt can point Hex's semantic models at those definitions.

That model works when the main output is an answer, report, dashboard, or internal data application. It becomes less natural when the work expands into several notebooks, local development, custom environments, model training, inference, or workflows that need to operate as part of a larger technical project.

Hex is a good fit for:

• business intelligence and reporting as the primary output
• stakeholders who consume dashboards rather than build analysis
• teams already invested in dbt MetricFlow, Cube, or Snowflake Semantic Views

Deepnote vs Hex pricing

Deepnote pricing has three plans.

• Free supports up to three editors and five projects, limited Deepnote AI, unlimited Basic machines, and seven-day revision history.
• Team costs $39 per editor per month when billed yearly. It includes unlimited viewers and notebooks, premium integrations, background execution, schedules, stronger machines, and monthly AI, CPU, and GPU credits.
• Enterprise adds permission groups, SSO and directory sync, private Docker images, bring-your-own LLM, audit logs, federated authentication, single-tenancy, and volume compute discounts.

Hex pricing has four plans.

• Community is free and includes up to five notebooks, small compute, and a Notebook Agent trial.
• Professional costs $36 per editor per month. It is positioned for individual users and includes unlimited notebooks, up to five published apps, the Notebook Agent, and 30-day version history.
• Team costs $75 per editor per month. It adds Threads, the Semantic Model Agent, unlimited apps, scheduled runs, Components, Collections, unlimited history, and advanced-compute add-ons.
• Enterprise is custom and adds audit logs, OIDC SSO, OAuth database connections, and optional single-tenant, HIPAA, embedding, and custom-image features.

The $36 Hex Professional plan is not the closest equivalent to Deepnote Team. Hex reserves its team collaboration, Threads, semantic-agent, unlimited-app, and scheduling features for the $75 Team plan.

Deepnote includes unlimited viewers, so the number of people reading notebooks, dashboards, or apps does not increase the seat bill. Hex separates author, Explorer, and Viewer roles, with Viewer billing depending on the plan.

Enterprise pricing is custom on both platforms. Both offer single-tenant deployments, but the deployment options differ: Hex's single-tenant VPC is fully managed by Hex, while Deepnote also supports customer-managed single tenancy and private-cloud or on-premise deployment.

For a data team comparing like-for-like paid plans, Deepnote has the lower editor price.