Product updates

Steep lets your whole organization work from the same trusted metrics and semantic model defined by your data team. The Steep MCP extends that power beyond Steep, into the AI tools and agents you already use every day.

Your data, where you work

Connect any MCP-compatible client to your Steep workspace and query your metrics directly from Claude, Cursor, Codex, VS Code, and more. Browse the metric catalog, query metric and target data, and explore your workspace without switching context.

Governed by default

Every query runs through the same semantic layer and permissions model that powers Steep. Members connect with their own Steep credentials via OAuth, so the AI sees exactly what the member is allowed to see, nothing more. Your data team stays in control. Also works with row-level security.

Getting started

The Steep MCP server is hosted at https://mcp.steep.app/mcp. Authentication is handled through OAuth, and your agent will only see data you have access to in Steep.

Learn more →

Row-level security (RLS) lets workspace admins control which rows of data each member can see.

Define who sees what

Admins can now set access policies directly on any module in Steep's semantic model. A policy filters the data a member sees based on their identity, so an account manager only sees data for their own accounts, or a regional lead only sees data for their market. The policy applies automatically across every metric, entity, and report built on that module.

User-level and custom attribute access

Policies can be set based on individual user email or custom attributes. Define a table mapping each member's email to their attributes, then set policies to filter by those attributes. Policies work with any dimension in your data, whether that's market, region, manager email, or any custom attribute.

Always consistent, always filtered

Once a policy is in place, members always see filtered data, whether they're exploring a metric, viewing a report, or using Steep AI. Members are informed when they're seeing a filtered view. Admins always see everything.

Availability

Row-level security is available on Enterprise plans.

Learn how to get started →

The shift to AI analytics

For years, getting data into the hands of the business has been a challenge for data teams. But in the AI era, access alone isn’t enough. Teams need the speed, flexibility, and depth of AI, but in an environment that stays trusted and grounded in the right business context.

Steep AI solves that problem.

Real analysis, not just answers

Built on top of your semantic model, the Steep agent gives teams a fast, powerful way to do analysis in natural language. It can connect metrics and dimensions, compare segments and time periods, and return clear answers grounded in the business logic your data team has defined.

• Ask complex questions in plain language and refine the answers in the same conversation.

• Explore your catalog using all metadata and descriptions available.

• Get deep analysis using the full context of your semantic model, with the Steep agent combining multiple metrics and breakdowns for real insights.

Read the full blog post →

Start using Steep AI →

With the Steep API, teams can access the metrics and semantic model defined in Steep and use them anywhere, while governance remains centralized. Complete consistency without the lock-in.

Unlock a whole new set of automation and integration use cases for analytics teams, whether you’re driving internal workflows, notebooks, or powering AI experiences with your metrics.

API for metadata and queries

This release focuses on the semantic model metadata and metric queries, providing the core building blocks for secure, read-only access to Steep’s semantic layer from other systems.

What’s included:

• Metadata endpoints to list metrics, entities, modules, members, and teams.
• Metric query endpoint that uses Steep’s query engine to compute and deliver aggregated metric data.

What this unlocks

Steep’s API opens up new possibilities for your analytics, without re-implementing definitions or logic.

• Build anywhere: Bring Steep’s metrics into your internal tools, notebooks or other systems.
• Introduce metrics to AI: Power LLMs, agents, and custom AI use cases with your trusted, governed metrics.
• Get creative: Build custom use cases built on top of Steep’s semantic model.

Explore the API documentation and references to get started:

https://help.steep.app/api

You can now define your semantic model in code. This unlocks a new level of control and sophistication for governance in Steep. Whether you define metrics in code or in the app, they’ll always work seamlessly side by side.

• Version control everything: track changes, review via pull requests, and keep a full history log of updates.
• Collaborate more efficiently: developers and data teams can safely iterate.
• Reuse definitions in multiple workspaces as a single source of truth.

Define anywhere

Define metrics, dimensions, and modules in code, or continue creating them directly in the Steep app, or use a mix of both. Choose the workflow that fits your team’s needs, and switch seamlessly between approaches.

This hybrid workflow combines the speed and flexibility of the app with the control, structure, and versioning of code, giving your data team the best of both worlds.

How it works

A powerful aspect of defining your metrics in code is that you can start small with just one module and build from there.

1. Define your modules, metrics, dimensions, and join paths in YAML
2. Connect your Steep workspace to a GitHub repository under Model → Code.
3. Push updates to your main branch. Steep validates the files and updates the model instantly, refreshing the metric cache automatically.

You can easily export in-app defined metrics to code-based definitions. Once synced, your module will be automatically overwritten and stay synced with your code.

Resources

• Learn more about define in code →
• YAML code reference →

---

Modules

Modules are replacing what we previously called the Semantic graph.

Modules are the semantic representation of your database tables. They give a single, unified view of the join paths, dimensions, metrics, and entities built from a table in your database. Now you can see how your tables are used across Steep, all in one place.

They also lay the groundwork for a well-structured semantic model that can be managed in code and enriched with AI-driven semantics.

No action needed

If you’re using Steep, your existing Semantic graph has been automatically migrated to Modules. You can find the Modules page under Settings in the Steep app.

Learn more about modules →