The Magnolia Agent: an open framework for agentic content operations

Search is changing, and so is the way people find your brand. Fewer people click through a page of blue links. More people ask an AI assistant for an answer and act on whatever it tells them. That shift changes what a website has to be. It's no longer enough to be a well-designed storefront for human visitors — your content also has to work as a structured, machine-readable source of truth that an AI model can read, interpret, and cite. If it can't, your brand is effectively invisible to a growing share of the people trying to find you.

That shift touches everyone building and running digital experiences. For marketers, it means content has to be created and optimized for two audiences: people and AI models. For developers, it means the systems behind that content need to expose their structure and meaning in ways an AI system can actually use — not just render nicely in a browser.

We built the Magnolia Agent to close that gap. It's a single AI agent designed to work inside your existing content operations, not next to them.

Most teams don't need another isolated AI feature bolted onto their CMS. They need help with the actual work: finding out what content is missing, creating it, getting it into every market that needs it, and making sure it performs once it's live. The Magnolia Agent is built around exactly that flow, in four connected stages: detect, create, scale, and optimize.

Planning and strategy. The Agent uses Context Search — the semantic search capability built on Magnolia's vector database — alongside Find Old Pages to compare what you have against what you should have and surface stale content for review. Point it at a new analyst report or a competitor's page, and it identifies the content gaps that matter, in minutes instead of a morning spent digging through Drive.

Creation and production. Once you know what to write, Doc-to-Page turns a brief or a document into a structured, on-brand page — including layout and components — instead of a blank canvas. It reuses the same vector database from the planning stage to select the right supporting assets from your existing library automatically; automatic image generation is on the roadmap toward GA.

Publishing and distribution. Getting a page live in one market is one thing. Getting it live, on-brand, and locally accurate across every market you operate in is another. The Agent's AI translation is page-structure aware: it understands headings, components, links, and calls to action, not just raw text, so layout and design stay intact as content moves across languages and out to every market that needs it.

Optimization and maintenance. Once a page is live, the Agent keeps it performing by refreshing titles and descriptions, structuring content so AI systems can understand and reuse it, and using real performance data to flag pages that need attention instead of relying on guesswork.

Each stage names real, working tools inside the Magnolia Agent today. If you're a marketer, this content lifecycle will look familiar — it's what you already do, just faster. (You'll also see this same lifecycle described as detect, create, scale, and optimize in From vision to workflow: the Magnolia Agent in daily marketing operations — same four moves, the vocabulary from the stage talk it's drawn from.) If you're a developer, those tool names are your entry point into the Agent's tool architecture — including the MCP tools that let it talk to external systems.

A lot of AI agents on the market today are closed systems: one vendor, one model, a fixed set of capabilities you can't change. We built the Magnolia Agent differently, because "practical" and "locked in" shouldn't have to go together.

The Agent is model-agnostic: choose the model that fits your standards, your budget, or your compliance requirements, through Magnolia's Unified Model Registry, and change that choice later without rebuilding anything. That matters more than it might sound — especially for enterprises that have already validated a specific model for security and compliance.

Model choice is one part of it. The other is extensibility: Partners and customers can write their own tools for the Agent — a workflow specific to a regulated industry, an integration with an internal system, a check that only makes sense for one team.

The Magnolia Agent is designed to grow with your operations, not to box you into ours. We'd also rather be precise than claim more than we can back up: real data sovereignty comes from running your own model on your own infrastructure. What the Agent guarantees today is choice, not lock-in — you're never stuck with a single provider, a single region, or a single cost structure.

The Agent isn't built on one closed integration layer. Its tools plug in through four paths:

MCP capabilities were added deliberately, on top of that foundation, precisely because of what an open standard buys you over time.

Tools registered through MCP work in both directions — the Magnolia Agent can call out to external MCP servers, and it can expose its own tools to external MCP clients — so the capabilities you build for the Magnolia Agent aren't trapped behind a single interface. It also means Magnolia DXP's AI investment extends past the Agent itself: the same open standard underpins the Magnolia MCP Developer's Server, which brings context-aware AI assistance to developers working on Magnolia projects. We go deeper on the Agent's tool architecture, including MCP, in Inside the Magnolia Agent: an open, extensible architectureMagnolia Agent: an open, extensible architecture.

Magnolia Agent is built from four named components that work together:

Mgnolia Agent ships under two consumption paths:

The Magnolia Agent is now generally available, model-agnostic across all providers.

From vision to workflow: Agent in daily marketing operations follows a fictional marketing team through detect, create, scale, and optimize — the version of this lifecycle you'll recognize from our NEXT26 talk — from a raw analyst report to a published, localized, optimized page.

Inside Magnolia Agent: an open, extensible architecture is the developer's view: how the agent loop, tool registry, and MCP actually work under the hood, and what it takes to build your own tool.

Doc-to-Page: from brief to on-brand page in minutes is a closer look at the single capability that keeps coming up in every demo — how a document and a layout choice become a live, structured page.

Introducing the Magnolia Vector Database for semantic search covers the context layer that will ground Agent in your own governed content — still in development, covered here for when it's ready.

Magnolia DXP and the EU AI Act lays out, plainly, what the Act's transparency requirements mean for Agent and for you, ahead of the Article 50 deadline.

Agent in practice: a practitioner's guide to the content lifecycle walks through the documented use cases stage by stage, for anyone who wants the hands-on reference rather than the narrative.