What Hex does
Hex combines collaborative SQL and Python notebooks, analytics agents, semantic context, conversational exploration, and publishable data applications.
Hex is an end-to-end analytics workspace rather than only an AI assistant. Data practitioners query with SQL, analyze with Python, visualize results, collaborate in notebooks, and publish the same project as an interactive app. Notebook Agent can help build and edit cells, while Threads and semantic-model agents extend natural-language access for broader teams. This continuity makes analytical logic reusable: a deeper investigation can become a governed application instead of being copied into a separate dashboard tool.
Trust depends on context and inspectability. Hex can combine database descriptions, semantic models, business rules, project logic, and curated guidance, and data teams can observe questions and improve that context. Users should still open generated code, confirm the connection and cache, test joins and grain, reproduce important numbers, and distinguish correlation from causation. An attractive generated app does not validate the analysis underneath it, and broad self-service requires deliberate data and project permissions.
Community is free with five notebooks and a Notebook Agent trial. Professional is $36 per editor monthly; Team is $75 per editor monthly and adds Threads Agent, Semantic Model Agent, extended credits, scheduled runs, and stronger collaboration. Enterprise is custom for capabilities such as OIDC SSO, audit logs, OAuth connections, single tenancy, HIPAA, embedding, and premium credits; advanced compute can be metered separately. Hex states SOC 2 Type II, encryption, ephemeral project data, and AI opt-outs, but notes that certain advanced model providers may retain prompts and outputs briefly for safety monitoring.
How Hex works
Hex connects to a data source and runs SQL, Python, chart, or no-code operations in an isolated notebook kernel. Agents can use workspace context, semantic models, project logic, and permitted data to draft or edit analysis, while Threads lets users explore conversationally and apps package approved work for others. Every result remains dependent on credentials, context quality, generated code, cache state, and review.
Bring governed data into a project
Hex authenticates to an approved warehouse or other source, applies workspace and project permissions, and runs queries against live data. Results enter an isolated notebook kernel and may use configurable caching; credential scope, upstream row security, cache age, and compute profile affect every later answer.
Develop analysis with code and agents
Analysts combine SQL, Python, charts, text, inputs, and reusable components, while Notebook Agent can draft or edit cells using permitted project and workspace context. Generated code is executable analysis, not proof: inspect packages, joins, units, missing values, sampling, and computational cost before accepting it.
Publish apps or conversational access
Approved notebook logic can become an interactive data app, scheduled result, Thread, Slack answer, or MCP-enabled workflow depending on plan and configuration. Semantic models and curated context help business users ask natural-language questions without exposing unrestricted editing or underlying credentials.
Observe answers and govern change
Reviewers reproduce key numbers, trace code and data lineage, test app controls, and verify that the narrative matches the analysis. Administrators monitor agent questions, context gaps, AI credits, compute charges, schedules, cache behavior, permissions, provider retention exceptions, versions, and audit logs before expanding access.
How to set up Hex
Choose the plan and cost boundary
Estimate editor and explorer roles, AI credits, published apps, schedules, advanced CPU or GPU compute, security add-ons, and enterprise support before committing.
Create a secure workspace
Configure identity, groups, project roles, AI settings, training opt-out, region, and retention; use Enterprise controls where organizational policy requires them.
Connect one governed data source
Use least-privilege credentials, restrict connections to approved users, verify warehouse row access, and document query cost, refresh, and caching expectations.
Build and test an analysis
Combine SQL, Python, charts, and agent assistance, then inspect generated code, reconcile key totals, test edge cases, and version the approved logic.
Publish with observability
Expose only necessary controls and data in the app or conversational surface, assign owners, and monitor runs, questions, costs, permissions, corrections, and context gaps.
Hex FAQs
Is Hex free?
Yes. Community is free and includes up to five notebooks, small compute, all cell types, data connections, and a Notebook Agent trial.
How much do paid Hex plans cost?
As reviewed July 15, 2026, Professional is $36 per editor monthly and Team is $75 per editor monthly. Enterprise pricing and several add-ons are custom.
What happened to Hex Magic?
Hex documentation says its AI features were formerly called Hex Magic. The current suite includes Notebook Agent and, on eligible plans, Threads and semantic-model agents.
Does Hex store all queried warehouse data permanently?
Hex describes project data as ephemeral in isolated kernel memory, with optional configurable caching. Credentials, uploads, cached results, metadata, and other service data have separate handling.
Do Hex model providers train on customer data?
Hex says its LLM providers do not train on customer data and generally use zero retention, but certain advanced models may retain prompts and outputs briefly for safety monitoring. Admins can disable AI and adjust training settings.
Listing reviewed 2026-07-15. Product details and pricing can change; verify important terms on the provider's website.
Related Research AI tools
Related AI guides
Reviews
Tell the community what you made, what worked, and what you wish you knew before starting.