Adobe Marketing Agent MCP Server

Switch to a different organization by exchanging the current token. Provide the organization name or IMS Org ID. Use list_orgs to see available organizations first.

This core system tool updates the current sandbox and/or dataview context for the session. Parameters: - sandboxName (optional): Name of the sandbox to set as current context it's usually one word - dataviewName (optional): Name of the dataview to set as current context The tool will: 1. Fetch the ID for each provided name from Adobe APIs 2. Update the session store with the IDs 3. Return confirmation of what was updated Note: At least one parameter must be provided based on what you want to change (sandbox/dataview). Use list_dataviews or list_sandboxes to see available options first.

Set the current sandbox context. Use this when the user wants to switch to a different sandbox. The response contains the displayValue (e.g., 'Prod - Production (VA7)') which should be used when talking to the user.

Set the current dataview context. Use this when the user wants to switch to a different dataview.

Get task status and events for an async task. Use cursor to get only new events since last poll.

List async tasks for a conversation context_id.

Submit user feedback about their experience with the AI assistant. Accepts a query parameter with the user's feedback message. Automatically classifies sentiment and submits to the feedback API. Use core-feedback-widget instead if the user wants to see a feedback form.

Read or clear the current user's persisted preferences (sandbox, dataview, org, region). Preferences persist for 90 days across sessions and surfaces. Set automatically when the user configures a sandbox or dataview. Parameters: - action (optional): "get" (default) to read, "clear" to delete Returns JSON with sandbox_name, sandbox_id, dataview_id, dataview_name, org_id, region, updated_at. Fields are null when not set.

Submit the user's final decision to complete or cancel a pending plan. Use this tool when the system has proposed a plan step that requires user confirmation before it can be executed — such as creating an audience, a journey, a campaign, or any other entity. This is the final step that completes the plan. ALWAYS use this tool when: - The system is waiting for the user's decision to proceed with a plan - The user provides an explicit approve or reject response to a pending proposal - User says "approve", "go ahead", "create it", "looks good", "confirm", "do it" - User says "reject", "cancel", "don't create it", "decline", "stop" Parameters: - decision (required): "approve" to confirm and execute the plan, "reject" to cancel it Returns the final orchestrator response confirming the plan was completed or cancelled.

This MCP server offers a comprehensive suite of capabilities designed to enhance user interaction with Adobe Experience Cloud applications. It enables users to discover and map XDM attributes relevant to their queries, analyze business metrics through visualizations and data tables, and gain insights into product features and best practices. Additionally, it provides statistical analysis of audience sizes and trends, evaluates journey performance and conflicts, and retrieves detailed metadata about various AEP entities, facilitating informed decision-making and operational efficiency.

Interactive widget to view and manage user context. Use this to display current organization, sandbox, and dataview, and to allow the user to switch between them. Call this widget when the user wants to see their context, list available sandboxes, list available dataviews, list available organizations, change sandbox, change dataview, or switch organization. Note: Users refer to sandboxes by title (e.g. "Production"), not by name (e.g. "prod").

Call this widget when the user wants to provide feedback or rate their experience. This displays an interactive feedback form with thumbs up/down buttons and rating categories. Users can select sentiment (like/dislike) and provide additional comments. If widgets are not supported by the surface, the system will automatically fall back to text-based feedback using core-provide_feedback tool.