Model Context Protocol (MCP)
Connect compatible AI agents and clients to your governed Stackless workspace.
Stackless provides a Model Context Protocol (MCP) server for exploring and operating the data assets available to your Stackless account from AI clients such as Claude, ChatGPT, Cursor, Claude Code, and Codex. The server uses your existing Stackless identity and permissions; connecting an AI client does not grant additional access.
Before you connect
You need:
- An active Stackless account.
- The hostname of your Stackless workspace, such as
https://your-workspace.stacklessdata.com. - MCP enabled for your workspace. If a client cannot reach the endpoint, ask your Stackless contact to enable it.
- An MCP client that supports remote Streamable HTTP servers with OAuth, or a terminal agent for the local connection.
Your MCP endpoints are:
| Client | Endpoint |
|---|---|
| Claude and most MCP clients | https://your-workspace.stacklessdata.com/api/v1/mcp/transport |
| ChatGPT | https://your-workspace.stacklessdata.com/api/v1/mcp/openai/transport |
Replace your-workspace with the hostname supplied by your Stackless
administrator. Do not paste access tokens, session cookies, or other secrets
into the connector URL.
What the connector can do
The tools your client sees depend on your Stackless role; tools you do not have permission for are not listed at all. Tool groups include:
| Area | Examples |
|---|---|
| Catalog and warehouse | Find relevant data, explain assets and columns, trace lineage, check freshness, and run read-only warehouse queries. |
| Sources | Inspect connectors and their health, start a new connection, sync, pause or resume, change schedules, and plan or apply schema changes. |
| Runs and operations | List and diagnose runs, read logs, cancel runs, and manage Operations failure alerts. |
| Transformation Models | Draft, preview, fork, and inspect models. |
| Semantic Models | Draft, validate, preview, and update models. |
| Dashboards | Find, inspect, draft, edit, validate, and preview dashboards, and manage scheduled exports. |
| Lifecycle | Review a draft's diff, publish, refresh, unpublish, or delete drafts. |
| Support | List your support tickets and read their conversations. |
| Monitors | Draft, test, and publish Monitors, where enabled for your workspace. |
Try prompts such as:
- “Find the data assets related to revenue and explain which one I should use.”
- “Trace the lineage of this dashboard and identify its upstream models.”
- “Show the status of recent Stackless runs and explain any failures.”
- “Which connectors are unhealthy, and why?”
Next steps
- Connect an AI agent or client.
- Review permissions and security.
- Use the troubleshooting guide if a connection or tool call fails.