MCP server

The DX MCP (Model Context Protocol) server acts as a bridge between your DX instance data and any MCP-compatible AI client such as Claude, Cursor, and Devin Desktop. Connecting it lets your agent query your DX dataset, explore your software ecosystem, and act on failing Scorecard checks directly from the client you already use. It’s available as a remotely hosted service or as a locally run server.

Looking for create/update/delete operations?
The MCP server is intentionally read-only. For enabling your AI agent with more DX capabilities — including creating scorecards, managing entities, creating studio reports, and more — connect it with the DX CLI instead.

Remotely hosted server

The remotely hosted MCP server is available at:

https://ai.getdx.com/mcp

Authentication

Authenticate your MCP requests by providing a bearer token in the Authorization HTTP header. This token can be either an organization token or personal access token.

Authorization

MCP tools only have access to those resources covered by the token’s scopes. For example, the queryData tool requires the datacloud:query scope. A couple of additional details:

• Revocation takes effect immediately. MCP requests are stateless, so once a token is deleted, subsequent requests fail immediately. Access is not retained for the remainder of a long-running session.
• Scope changes apply immediately. Because scopes are evaluated at tool call time (see Authorization), any change to a token’s scopes — or to the scopes the user’s role allows — is enforced on the next tool call.

Configuration

Claude Code

claude mcp add --transport http dx-mcp https://ai.getdx.com/mcp --header "Authorization: Bearer [TOKEN]"

Cursor

Add the following to your Cursor MCP configuration:

{
  "mcpServers": {
    "dx-mcp": {
      "url": "https://ai.getdx.com/mcp",
      "headers": {
        "Authorization": "Bearer [TOKEN]"
      }
    }
  }
}

Local deployment

If you prefer to run the MCP server locally instead of using the hosted version, you can follow the instructions on the dx-mcp-server GitHub repository.

Rate limiting and audit logging

Because the MCP server utilizes the DX Web API underneath, tool calls are subject to the same limits and are recorded in the audit log just like direct Web API requests.

Available tools

Entities

Tool	Required scope	Description
listEntities	catalog:read	List entities from the DX software catalog, with optional search and type filtering.
getEntityDetails	catalog:read	Get full details for an entity including its metadata, tasks, and scorecard results.

Scorecards & initiatives

Tool	Required scope	Description
listScorecards	scorecards:read	List all active scorecards.
getScorecardInfo	scorecards:read	Get details about a specific scorecard, including its levels and checks.
listInitiatives	scorecards:read	List all initiatives with summary information.
getInitiativeDetails	scorecards:read	Get initiative details along with its progress report across entities.
reviewTasks		Given an entity and a set of check IDs, produces a structured plan or executes fixes for failing scorecard tasks.

Teams

Tool	Required scope	Description
listTeams	snapshots:read	List all teams in DX.
getTeamDetails	snapshots:read	Get details for a specific team by ID, reference ID, or member emails.

Data

Tool	Required scope	Description
queryData	datacloud:query	Execute a SQL query against the DX Data Cloud PostgreSQL database. Use information_schema when table/column structure is uncertain.

Note: The datacloud:query scope also requires Data Studio access in addition to the scope itself.

---

Available prompts

Prompt	Description
reviewTasks	A guided prompt that walks an AI agent through identifying, analyzing, and resolving failing DX scorecard checks for an entity.