At Adobe Summit 2026, Adobe announced the Marketo Engage MCP Server — a native Model Context Protocol endpoint that lets AI agents like Claude, GitHub Copilot, Cursor, and Gemini interact with a Marketo instance safely, without burning API rate limits. It entered Private Beta in Q2 2026, and moved to Open Beta by May 2026. For marketing operations teams, this is not a nice-to-have novelty. It is the beginning of a new operating model.
What the Marketo MCP Server actually is
MCP — the Model Context Protocol — is an open standard for connecting large language models and AI agents to external systems. Anthropic introduced it in late 2024, and it has since been adopted by every major AI vendor. Adobe's Marketo Engage MCP Server is a compliant MCP endpoint that exposes over 100 Marketo operations through a secure, role-aware bridge.
What operations are exposed
- Forms — read, create, update, retrieve fields
- Programs — list, create, clone, update settings and metadata
- Smart Campaigns — read flow steps, filters, triggers, schedule
- Leads / People — search, read, update, list by activity
- Emails — read HTML, update subject lines and tokens, list templates
- Snippets — create, update, and retrieve reusable content blocks
- Static Lists — create, add and remove members, list membership
- Folders — read structure and move assets across workspaces
Why this matters for marketing operations
Until now, most AI use inside Marketo has been narrow — an assistant inside a single feature, or an external tool bolted on with a REST integration. MCP flips that. Any MCP-compatible AI agent can now reason across your whole Marketo instance, take actions, and be governed by the same role-based permissions your users already have.
“Marketo becomes callable by AI agents rather than only usable by humans. That is a structural change, not a feature release.”
Three real use cases that are already viable
- Instance audit — an agent walks your program folder, flags naming convention violations, and produces a written report an operator can review before executing changes.
- Program QA — before a launch, an agent inspects tokens, list membership, sync direction, and channel-to-status mapping, then posts a checklist result to Slack.
- Content review — an agent reads every email in a campaign, checks against brand rules, and returns diffs the marketer can accept in-editor.
Governance and security
MCP usage does not consume standard Marketo REST API rate limits — that alone changes the economics of automation. Access flows through Adobe's authentication framework, so an AI agent is bound by whatever role the user connecting it has. If a user cannot delete programs in the UI, the agent cannot either.
The three deprecations you need to plan for in parallel
Adobe's July 2026 release notes bundle three API changes alongside the MCP rollout. Any team preparing for MCP should treat these as prerequisites:
- REST API access_token as a query parameter is unavailable after July 31, 2026. Integrations must use the Authorization header.
- SOAP API reaches end of life on July 31, 2026. Any legacy service still on SOAP must migrate to REST.
- Merge Leads API calls with more than 25 IDs will start returning error 1080 from July 31, 2026 and be skipped.
What we recommend doing now
The teams that will get the most out of MCP over the next 12 months are the ones that clean up their instance first. An AI agent inheriting a messy Marketo instance amplifies the mess. An AI agent inheriting a well-governed instance compounds the leverage. Start with a naming convention audit, a template rationalization pass, and a sync-health snapshot. Then request Open Beta access.
If your team is exploring Marketo MCP and wants a working session with senior operators who have shipped this pattern in other stacks, this is the kind of work we do at Xadentra.