Omni MCP Server
Query your governed Omni BI data in natural language from Claude, Cursor, and other AI tools.
Updated
What is Omni MCP?
The Omni MCP server is a Model Context Protocol server that lets AI assistants query your Omni business intelligence data in natural language, governed by Omni's semantic model and your own access controls.
Instead of traditional text-to-SQL — which can hallucinate or leak data — Omni routes AI queries through its semantic model so responses stay grounded in the business logic and metrics your team has already defined. The assistant picks a model and topic, then executes governed queries; results can be chained for iterative, multi-step analysis. A built-in documentation search tool can also answer questions about Omni itself with citations.
The remote server is hosted by Omni at https://callbacks.omniapp.co/callback/mcp and authenticates over OAuth 2.1 (with PKCE). On first connect, your browser opens an Omni authorization page; after you approve, Omni automatically issues a personal access token tied to your account, so every MCP query inherits your in-app Omni permissions.
Tools the Omni MCP server exposes
| Tool | What it does |
|---|---|
| pickModel | Returns the list of available Omni models and their IDs to query against. |
| pickTopic | Returns the available topics within a selected model to scope the query. |
| getData | Executes a query against the selected model and topic (or view) and returns governed results. |
| searchOmniDocs | AI-powered search of Omni's documentation that synthesizes answers with citations to specific pages. |
Connect the Omni MCP server
Claude Code
- 1
Add the server
Run: claude mcp add --transport http omni https://callbacks.omniapp.co/callback/mcp
- 2
Authenticate
Run /mcp, select the Omni server, and choose Authenticate. Your browser opens the Omni authorization page.
- 3
Authorize
Review the requested permissions and click Authorize. Omni creates a personal access token linked to your account automatically.
- 4
Query
Ask data questions in natural language; the agent uses pickModel/pickTopic/getData to return governed results.
Cursor
- 1
Add the MCP server
Open Cursor Settings > Tools & Integrations > New MCP Server.
- 2
Configure mcp.json
Add: { "mcpServers": { "Omni": { "url": "https://callbacks.omniapp.co/callback/mcp" } } }
- 3
Connect via OAuth
Click Connect next to the Omni server; your browser opens the Omni authorization page.
- 4
Authorize
Review permissions and click Authorize. Omni issues a personal access token scoped to your Omni permissions.
Sample use cases
Run ad-hoc data questions in natural language without writing SQL
The agent translates the question into a governed query against your Omni semantic model and returns the answer.
Generate daily KPI summaries from governed data models
A scheduled agent pulls key metrics each morning and posts a summary to Slack or email.
Compare metrics across time periods for trend analysis
The agent queries the same topic for two date ranges and highlights the differences.
Pull data for weekly stakeholder reports
The agent queries multiple topics, formats the numbers, and drafts a report ready for distribution.
Validate a hypothesis by querying specific segments
The agent filters data by segment (region, cohort, product line) and returns the results for analysis.
Search Omni documentation for setup or configuration help
The agent uses searchOmniDocs to find answers with citations to specific documentation pages.
Audit data model coverage to find ungoverned metrics
The agent lists available models and topics, compares them against a checklist, and flags gaps.
Security & permissions
Authentication is OAuth 2.1 with PKCE. On connect, Omni opens a browser authorization flow and — once you approve — automatically issues an "MCP OAuth PAT" tied to your Omni account. All MCP queries then run under your in-app Omni permissions and access controls, so the agent can only read data you are already allowed to see; the semantic model further constrains responses to defined business logic. (An alternative API-key/header method exists for self-managed instances, but the hosted remote server uses OAuth.)
Using Omni MCP with Gamut
In an automated Gamut workflow, a Gamut agent uses the Omni MCP server as a tool to pull governed metrics on a schedule or in response to an event — for example, querying a revenue or product topic each morning, or fetching the latest figures when a webhook fires, then routing the answer into a report, Slack message, or downstream step. Gamut manages the OAuth connection to Omni so the agent can call the server without manual key handling, and because queries run under the connected user's Omni permissions, the agent only ever sees data that user is allowed to access.
Frequently asked questions
What is the Omni MCP server?
It is a Model Context Protocol server from Omni Analytics that lets AI assistants query your Omni business intelligence data in natural language. Queries run through Omni's semantic model and your access controls, so answers stay governed and grounded in your defined metrics.
Is the Omni MCP server free?
The MCP server is included with Omni Analytics rather than sold separately, so you need an Omni account to use it. There is no standalone free tier; check Omni directly for plan details.
How do I connect the Omni MCP server to Claude Code or Cursor?
Point your client at the remote URL https://callbacks.omniapp.co/callback/mcp. In Claude Code, run claude mcp add --transport http omni <url>, then /mcp to authenticate. In Cursor, add it under Settings > Tools & Integrations and click Connect.
Is the Omni MCP server hosted or self-hosted?
Omni hosts a remote MCP server at https://callbacks.omniapp.co/callback/mcp that uses OAuth 2.1. An API-key option for self-managed setups also exists, but the standard path is the hosted remote server.
Can the agent access raw SQL or only governed models?
The agent queries through Omni's semantic layer — it picks a model, picks a topic, and gets governed results. It does not run raw SQL, which keeps answers consistent with your defined metrics and access controls.
Does it respect my Omni role-based permissions?
Yes. The agent authenticates as your Omni user and inherits your role's data access. If your role restricts certain models or fields, the agent cannot query them.
Can I query multiple Omni models in the same session?
Yes. The agent can call pickModel to switch between models within a single session. Each query can target a different model and topic as needed.