MCP server approval workflow

  • Release version: Yokohama
  • Updated November 23, 2025
  • 1 minute to read

    • Explore the Model Context Protocol (MCP) server approval workflow.

    Before you begin

    Role required: AI steward [sn_ai_governance.ai_steward]

    Procedure

    1. Navigate to Al Control Tower > AI assets > Approvals.
    2. Select the MCP server from the asset type filter or navigate to Lifecycle and filter by the current state of New.
    3. Select the MCP Server record with the status of New.
    4. Select Start Review.
      Note:
      Within AI Agent Studio, the MCP Server shows a status of AI Steward review during this phase. Product Owners can select View approval record in AI Agent Studio to go directly to the MCP Server record in the AI Control Tower.
      A playbook is triggered to guide the approval process through these phases: Assess, Build and Test, and Deploy.
    5. Select Mark as Complete for the Assess phase.
      Note:
      The Asses phase includes the Architecture review, Risk Assessment, and Stakeholder review.
    6. Complete the Build and Test phase activities.
      Note:
      The Build and Test phase involves out-of-application testing and validation activities such as running a test AI agent using the MCP Server as a tool and validating the output received.
    7. Select Mark as Complete for the Build and test phase.
    8. Select Mark as Complete for the Deploy phase.
      Note:
      The Deploy phase is the final step to deploy and enable the MCP Server for usage.
    9. Select Save.
      Note:

      The AI Gateway setup tab appears only after an MCP server request is approved. The AI steward can pause transactions whenever needed through this tab. For information on the tabs appearing in the MCP server record, see MCP server record.