New: Adobe Marketo Engage MCP

Overview

You can now connect Workfront Fusion directly to the Adobe Marketo Engage MCP server using the new Adobe Marketo Engage MCP connector. The connector exposes an instruction‑driven module that lets you provide open‑ended, plain‑English instructions; an AI agent then determines and executes the Marketo actions required to fulfill your request.

What’s new

• Adobe Marketo Engage MCP connector — a new connector that talks to the Marketo MCP server and exposes Marketo tools to MCP‑compatible agents.
• Instruction module — enter free‑form, plain‑English instructions (for example, “Create a nurturing program for new trial users in EMEA and add a 3‑step email sequence”) and let the agent translate them into concrete Marketo operations.
• Action restrictions — admins can explicitly limit which read, write, and delete actions the agent is allowed to perform so unattended runs remain within the approvals you set.
• AI provider flexibility — by default the module uses Adobe’s managed AI service; you can optionally provide your own OpenAI, Anthropic Claude, or Amazon Bedrock key if you prefer to route requests to another model provider.
• Tooling parity — Marketo operations exposed through the MCP connector follow the same permission model and tool labeling (Read/Write) as other Fusion MCP connectors.

How it works

1. Create a Marketo connection: In Fusion, go to Connections > Create Connection and choose Adobe Marketo Engage MCP. Provide organization and authentication details for your Marketo MCP server.
2. Add the Adobe Marketo Engage MCP module to a scenario or use it via an MCP agent (Coworker, custom MCP client).
3. Enter a plain‑English instruction describing the Marketo task you want performed (for example, “Export all leads created in the last 7 days with UTM source=paid_search”).
4. Optionally set an action policy that restricts which Marketo read/write/delete operations the agent can invoke.
5. Choose the AI provider: use Adobe’s managed AI by default, or supply your own OpenAI / Anthropic / Bedrock API key to run instructions against your preferred model.
6. The agent analyzes the instruction, maps it to one or more Marketo API operations, and executes them with the permissions you granted. The module returns an execution summary and any output (export files, created assets, logs).

Why this matters

• Faster, conversational automation — non‑technical users can drive Marketo actions with natural language without building complex scenarios or custom scripting.
• Safe unattended runs — action restrictions let you run unattended automations while limiting the agent to only the operations you approve, reducing risk.
• Flexible AI choice — teams can use Adobe’s managed AI for a turnkey experience or bring their own model keys for governance, latency, or cost reasons.
• Integration consistency — the connector follows Fusion’s MCP tool and permission patterns so it fits into existing admin and security workflows.

How to get started

1. Open Fusion > Connections > Create Connection > select Adobe Marketo Engage MCP.
2. Configure connection details for your Marketo MCP server and save.
3. Add the Marketo MCP module to a scenario or open Coworker / another MCP client and connect to your Fusion organization.
4. Compose a plain‑English instruction and run it. Start with read‑only queries (for example, “List programs in the Marketing folder created in the last 30 days”) to verify behavior.
5. Set action restrictions in the connector or scenario if you plan to allow write or delete operations. Test with a small, non‑destructive instruction before scaling up.
6. If you want to use a non‑Adobe model, provide your OpenAI, Anthropic Claude, or Amazon Bedrock key in the module settings and save.

Troubleshooting (quick)

• Connection failures: Verify the MCP server URL, OAuth/client credentials, and that your Marketo organization allows MCP access. Check network reachability and firewall rules.
• Agent won’t perform an action: Confirm the requested action is included in the allowed actions list. If an operation is blocked, the module returns a clear denial and execution summary.
• Unexpected or incorrect output: Review the instruction for ambiguity. Try a more specific prompt (step‑by‑step or enumerated requirements) and re‑run. Include example data where applicable.
• AI provider errors: If using a custom model key, confirm the key is valid, has remaining quota, and the provider endpoint is reachable. Check provider error messages returned by the module for guidance.
• Large or long‑running jobs: For exports or bulk updates, prefer scenario‑driven runs with explicit retry/timeouts configured; monitor execution logs for progress and partial outputs.

Documentation

• See the Experience League reference for details and configuration steps: Adobe Marketo Engage MCP module (Experience League)

Feedback and support

Tell us how you’re using the Marketo MCP connector or request enhancements (examples: richer prompt templates, per‑action approval workflows, built‑in audit summaries, or automated rollback hooks) by:

• Posting in the Experience League Community thread for Marketo or the Fusion MCP connectors.
• Opening a support ticket via the Admin Console Support portal if you encounter unexpected behavior or need migration help.
• Including example instructions, expected vs actual results, and connection metadata when reporting issues to help us reproduce and diagnose problems faster.

We’re excited to bring conversational, policy‑controlled Marketo automation to Fusion. Try the new Adobe Marketo Engage MCP connector and let us know what prompts and controls you’d like next.