API Access (MCP)
Connect your own AI assistant — Claude, Cursor, ChatGPT, or any MCP-compatible client — directly to your Seemore data through the Model Context Protocol (MCP) server, or query the same data over a sta
Overview
Seemore exposes a Model Context Protocol (MCP) server so you can connect your own AI assistant — Claude, Cursor, ChatGPT, or any MCP-compatible client — directly to your Seemore account.
Model Context Protocol (MCP) is an open standard that lets AI assistants call external tools. Once connected, you can ask questions in plain language and your assistant will explore data assets and lineage, break down Snowflake cost and usage, investigate anomalies, review query history, run Seemore's AI optimization agents, and manage tags — all scoped to exactly what your account and role already allow.
The same data is also available over a standard REST API, which is ideal for scripts and partner integrations.
Required role to connect: Viewer or higher.
Any authenticated user can connect an AI tool, generate a personal access token, and read data through the API. The API only ever returns data your role and integration access already permit. Creating, editing, or deleting custom tags through the MCP requires Editor or higher.
Endpoints
You'll find your personal connection details under Settings → API Access.
MCP
Your personalized URL, shown under Settings → API Access (based on https://external-api.seemoredata.io/mcp)
REST
https://external-api.seemoredata.io/v1
OpenAPI
https://external-api.seemoredata.io/swagger
Your MCP URL is personalized to your organization. Always copy the exact URL shown in Settings → API Access rather than typing it by hand — federated SSO tenants receive a URL that routes sign-in through their identity provider.
What you can do
Once connected, your AI assistant can answer questions like:
"What were my five most expensive warehouses last week, and how does that compare to the week before?"
"Why did compute cost spike on Tuesday? Trace it to the root cause."
"Show me the lineage for the
ORDERStable and flag any downstream tables that are never queried.""List my open cost anomalies, then analyze the highest-severity one."
"Find the slowest recurring queries this week and suggest optimizations."
"Which budgets are over 80% spent this month?"
Behind the scenes, these requests map to a catalog of read-only tools for assets and lineage, cost and usage trends, warehouses, budgets, anomalies, insights, query history, and tags — plus two AI agents that stream live root-cause and query-optimization analysis. See the full list in the MCP Tool Reference.
How authentication works
There are two ways to authenticate, both managed from Settings → API Access:
Browser sign-in (recommended for AI tools) — Paste your MCP URL into Cursor, Claude, or ChatGPT and the tool opens a browser tab to sign you in to Seemore. There is no token to copy or store. Each connected tool appears under Connected AI tools, where you can revoke its access at any time.
Personal access token (for scripts & REST) — Create a named token under the Personal access tokens tab and send it as a
Bearertoken. Use this for scripts, partner integrations, andcurl.
See Connecting AI Tools for step-by-step setup instructions.
Security & scope
Your warehouse data is never modified. The API reads Snowflake metadata and usage that Seemore already collects. The only write operations are managing Seemore custom-tag definitions, which require the Editor role.
Scoped to you. Every request is limited to the integrations and data your Seemore role and access already permit. A connected AI tool can never see more than you can.
Revocable at any time. Revoke a personal access token or a connected AI tool from Settings → API Access and it stops working immediately on its next call.
Next steps
Connecting AI ToolsMCP Tool ReferenceLast updated