What Hex is
Hex is a collaborative analytics and data-application platform. Teams connect data, use SQL, Python, Markdown, charts, pivots, and visual exploration in notebooks, collaborate, schedule work, and publish interactive data apps.
Hex's AI features include a Notebook Agent for technical work, Threads for conversational self-service, a Modeling Agent for semantic models, Chat with App, and tools for generating data apps. The platform can move from analysis to an operational interface, which makes governance more important than in a standalone chatbot.
Notebook Agent and executable logic
The Notebook Agent can create and modify SQL, Python, Markdown, pivot, and chart cells. Hex documentation explicitly says it is intended for technical users who can audit the suggested SQL and code.
Use a development project and read-only data role. Inspect joins, filters, time zones, units, null handling, sampling, statistical assumptions, package dependencies, network calls, and output destinations. Add tests for known metrics and row-level permission boundaries.
Generated Apps are currently described as beta. Do not promote a generated application because its interface looks complete. Review every query, input, filter, permission, cache, export, error state, and disclosure of underlying data.
Threads, semantic models, and self-service
Threads lets Explorer users ask questions over curated context. The Agent prioritizes endorsed and semantically modeled data. A Modeling Agent can help create and edit those semantic models, while Context Studio lets administrators monitor usage and influence context.
The semantic layer is the control surface for business definitions. Assign owners to metrics, dimensions, joins, descriptions, freshness, and deprecation. AI-generated model changes should go through code or governance review before they affect self-service answers.
Restrict raw schema access. A user who can ask a question should not automatically see every row or underlying table. Test data permissions, app sharing, Explore behavior, exports, Slack, MCP, CLI, and employee offboarding.
Data apps, schedules, and compute
Hex can publish notebooks as interactive apps, schedule runs and alerts, embed analytics, and connect external interfaces. These capabilities create long-lived operational dependencies. Record owner, source freshness, schedule, recipients, credentials, failure handling, and rollback.
Medium compute is included on paid plans. Team and Enterprise can opt into advanced compute billed by the minute, with current profiles ranging from CPU sizes to GPUs. A generated workload can trigger expensive queries, repeated notebook runs, or high compute. Set runtime, warehouse, and cost limits.
Test concurrency, cache behavior, timeouts, package installation, environment isolation, secrets, exports, and app permissions. Published data must not reveal sensitive values through filters, downloads, tooltips, or generated summaries.
Current plans and seat cost
Hex pricing currently lists Community as free with up to five notebooks, an Agent trial, data connections, all cell types, and Small compute. Professional is $36 per Editor per month and adds Notebook Agent, standard credits, unlimited notebooks, up to five apps, 30-day history, and Medium compute.
Team is $75 per Editor per month and adds Threads, Modeling Agent, extended credits, unlimited apps and history, scheduled runs, shared components, and advanced compute add-ons. Enterprise is custom and can add Explorer seats, audit logs, premium credits, OAuth database connections, observability, OIDC SSO, single tenancy, HIPAA, embedding, custom images, and dedicated support.
Model Editor and Explorer seats, apps, credit demand, advanced compute, embedding, single tenancy, HIPAA, support, warehouse cost, and idle projects. A 14-day Team trial currently requires no payment card and falls back to Community if not upgraded.
AI credits and automatic top-ups
Hex currently grants 30 monthly credits to a Professional Editor, 40 to a Team Editor, 10 to a Team Explorer, 60 to an Enterprise Editor, and 10 to an Enterprise Explorer under standard seat subscriptions. Grants reset, cannot be shared, and do not roll over.
Agent tasks consume effort-based credits. Official examples range from roughly one credit for a simple count to around ten for a generated data app, but actual usage varies with data, context, history, and task. External Agent runs from Slack, MCP, or CLI also consume credits.
Administrators can purchase pooled add-ons. Auto top-up currently buys 50 credits for $25 when the pool drops below 25; admins can set spend and user allocation limits. Purchased auto-top-up credits roll over. Enable a cap before broad Agent access.
AI privacy, security, and plan controls
Hex says AI providers do not train on customer data and usually operate under zero-data-retention agreements. Some advanced models require temporary provider retention for safety and security. They are only available when an administrator enables them, and retained data is not used for training.
Enterprise can use supported BYOK, which routes model use through customer provider keys and avoids Hex credit consumption, while shifting cost and usage visibility to the provider. BYOK does not eliminate prompts, warehouse access, Hex logs, or other platform data routes.
Review the DPA, security program, subprocessors, regions, encryption, support access, retention, deletion, audit logs, SSO, single tenancy, HIPAA add-on, incident response, and prohibited sensitive data. Current terms say US customers should not upload PHI without a BAA.
Who should choose Hex?
Hex is a strong shortlist for data organizations that need collaborative code, governed self-service, semantic context, and published data applications in one platform. It is most appropriate when a technical team owns the warehouse and reviews AI-generated logic.
Choose Julius AI for a conversational file and data Agent with individual power-user plans. Choose Rows AI for spreadsheet-first analysis and business integrations. Choose Elicit for academic evidence workflows rather than operational analytics. The AI research and data analysis category provides a common code, permission, credit, compute, and privacy framework.
Visit Hex
