We built Omni around a governed semantic layer.
Define a metric in the semantic model, and every surface reads from it and returns the same answer, whether it's someone using AI or connecting to Omni through another tool. Make a change, and every dashboard, agent, and App benefits.
Spreadsheets
Dashboards & Apps
Point & click UI
Conversational Analytics
Embedded Analytics
SQL
MCP &Conversational AnalyticsPoint & click UISQLEmbedded AnalyticsDashboards & AppsMCP &Spreadsheets
Giving you governance without giving up speed.
Omni's semantic layer is:Explore first, then model what matters. Build from the browser, in code, or with AI. When you've created something valuable, promote it safely - branch, review, evaluate, and merge its definition into the shared model.
Bi-directional integrations with other semantic layers, such as dbt, Snowflake Semantic Views, and Databricks Unity Catalog. Written in YAML, committed to open standards (Apache Ossie), and portable by design.
Handle exceptions without duplication or workarounds. Extend the model at any grain with custom logic, AI settings, display defaults, and more. When a core definition changes, the additive logic updates automatically.
Every metric, join path, and edge case can be defined once. AI context teaches agents what your best experts already know.
Every query is compiled from the model, enforcing the same definitions, permissions, and row-level filters.
Business users at Synthesia define new metrics in Omni's UI, and the data team reviews and promotes them through the semantic layer. Expertise lands in the model without losing consistency.
Read the case study
Omni flips the traditional data model workflow. Instead of the data team coding every metric definition, business users can define them directly in the UI, using Excel-style formulas or point-and-click. The data team then reviews these definitions in a git workflow, and once they're ready, promotes those definitions across the company for reuse. It's how we move fast without sacrificing trust.
Edward Mancey, Director of Data, Synthesia
Omni's model is written in YAML, and committed to open semantics through Apache Ossie. Your data stays in your warehouse, logic stays version-controlled in Git, and integrations run in both directions, giving you a shared interface for your data stack.
The dbt integration is bi-directional. View dbt metadata in Omni, author models from Omni queries, and push logic upstream. Branch mode isolates the change, and git keeps the history.
Snowflake Semantic Views and Databricks Unity Catalog Metric Views can be imported and queried directly. Explore, add AI context, and version them in Omni, then push definitions back out with Omni's open source agent skills.
Native connections support Snowflake, Databricks, ClickHouse, and more.
Every company has exceptions. Regional rules, team-specific metrics, unique tenant needs. When the semantic layer can't hold them, people copy the model and route around it. Numbers drift, answers conflict. Omni's hub-and-spoke model handles the exceptions, while core definitions stay intact.
A central hub model holds the shared logic. A department or tenant creates a spoke that inherits all of it, and adds or overrides only what's different. Unique logic, permissions, and styling are handled without duplicating the base model.
Adding a region or tenant means adding a spoke, not a new model or duplicate dashboards. Dynamic shared extensions read each viewer's user attributes and route them to the right spoke automatically.
Each person sees their own metrics and only the data they're allowed to see. Domain-level permissions control visibility, and Git followers keep every deployment on the same version of the model.
Read the case studyThe semantic layer is deterministic and precise; it defines what calculations exist. We use Omni's AI context to be conversational and contextual, to help AI understand how we talk. We don't duplicate across the layers; we use them to translate so that Omni's agent can have a real conversation with someone who understands our business but doesn't write SQL.
Sarah Fischbach, Staff Analytics Engineer, Checkr
Your skepticism about AI on company data is earned. AI by itself doesn't fix chaos, it amplifies it. Omni fixes this: every question, from every surface, runs through one governed model, and returns a reliable answer. And that governance doesn't cost you speed.
Whether the question comes from a dashboard, workbook, or AI chat, it returns the one answer your team defined. Access filters enforce row-level security, and access grants control columns and datasets. That consistency earns trust.
The fix for drifting answers is architectural, not a better prompt or AI model. Omni's AI harness enforces guardrails: a question routes to one Topic, the agent selects fields and filters, and our query engine compiles the SQL.
The model governs what people see. The platform protects the rest. Omni is ISO 27001 certified and SOC 2 Type II audited annually and meets GDPR, CCPA, and HIPAA standards. Customer data is encrypted with AES-256 at rest and TLS in transit, and SAML controls sign-on.
Read the case studyCreating the Omni data model was very fast compared to what we were used to. The generation of the initial model code in Omni is so easy — instantly syncing with our warehouse and having views available saved us a lot of time. We built a lot of curated datasets (Topics) and had dashboards ready in a couple of weeks.
Lizzy Bradford, Senior Director of Analytics, BuzzFeed
