In Part 1 of this series, we wanted to give Marketing Engineers the ability to combine their AI tools, synthesize insights across them, and have the flexibility to do that both inside Profound and out in whatever tools they already live in. The first step was surfacing the same analytics you get in Profound via the MCP so that Claude could work with that information.
Today, we're taking the next step. Profound is now available as an official Anthropic connector, which means you can pull Profound into Claude in a couple of clicks. There’s no setup or config file required.
But the connector is the easy part to announce. The real story is what capabilities this unlocks: the ability to pull data and push actions back and forth between Profound and Claude.
New capabilities unlocked with the Profound connector
Part 1 was about answers. You could ask your agent how your visibility moved last week and get a contextualized response instead of a data dump. That's helpful, but it's still just a read. You'd get your answer, then tab back into Profound to actually do something about it. Phase 2 closes the loop.
Generating prompts to add to Profound
You can now update the prompt set that you're tracking in Answer Engine Insights by simply telling Claude to "Add these ten prompts to my enterprise CRM category." "Generate prompt variations around our new pricing page." Claude calls the right capability, the prompts are updated in your Profound instance, and you never need to leave the conversation.
Build and run a Profound Agent from inside Claude
Design, build, and run Profound Agents directly in Claude. Think of a Profound Agent as a marketing specialist. They're composed of building blocks that all come together to accomplish a marketing task. Agents can synthesize deep insights from your Profound data, produce AEO content, or identify content gaps.
Now, you describe what kind of Agent you want in plain language inside Claude, and it gets designed and is ready to run in the same session.
You can even work with the MCP, directly in conversation, to get ideas about workflows that can be automated using Profound Agents. Think of Claude as the orchestrator and a Profound Agent as a specialist with a targeted job. More on this in the next section.
It's not Claude talking to Profound. It's Claude building inside Profound on your behalf. And because every agent runs on the same runtime that powers AI Search efforts at scale for thousands of our enterprise customers, what you stand up in a sentence is a reusable Agent that can be shared across teams.
A generalist orchestrating specialists
There's a useful way to think about what's actually happening here. Claude is a generalist. It's broad, capable, and great at reasoning across whatever you throw at it. But, it isn't a marketing expert, and it shouldn't have to be. A Profound Agent is the opposite: a narrow marketing specialist that does one marketing job extremely well on top of your brand's marketing context, and our data and runtime. Put them together and you get an orchestrator and its sub-agents, a mixture of experts.
For marketers building entirely in Profound, Aim is the latest version of our orchestrator that is constantly scanning your data and collaborates with you to build Agents in Profound. But for those who choose to work in Claude, Claude is the orchestrator. Through our MCP, it calls on Profound Agents that carry what it can't: your tracked prompts, your competitive set, your visibility and citation data, the AEO methodology behind every metric. You ask one question in one session; Claude figures out which specialist to call, runs it, and brings the result back into the conversation.
What's next: More specialists coming
Today you can build and run a marketing agent through the MCP. The orchestrator-and-subagents pattern is already live: Claude coordinates, a Profound Agent executes. What's coming next is range. Right now the specialists you can spawn cover a focused set of jobs. The roadmap is to widen that catalog: more types, more diversity, more depth of marketing agent you can describe in a sentence and put to work. The richer the bench of specialists Claude can reach for, the more of your AEO work moves from "open the dashboard and do it" to "ask, and it's done."
When we wrote about Cortex, we said we were building for a world where autonomous AI systems, not dashboards, are the primary consumers of analytical tools. Every new specialist we add through the MCP is a step into that world. That's where the next post in this series is headed.
Get started
The Profound MCP is available today for Enterprise customers, with setup instructions here. Self-serve customers and others interested are encouraged to get a demo to see it in action.
