Introducing: Outset Agent

Most research questions don't start in a research tool.

They start in Slack. In a Figma file with a comment from a designer. In the Notion doc where the brand team is debating positioning. By the time a question reaches a researcher, it's already circulating in three other places and the team has usually made a rough call without waiting.

The problem isn't that teams don't value research. It's that running it still requires everyone to stop what they're doing and go somewhere else. Open a new tab. Log in. Build the study. Launch it. Then go back to wherever the actual work was happening.

We've been thinking about that gap for a while.

What we shipped

Today we're releasing the Outset Agent that powers two things.

Outset now connects to Claude, ChatGPT, Cursor, Figma, Notion, Glean, and the other tools your team already works in. Tell your agent what you need to learn - it creates the study, configures recruitment, launches your study, and analyzes the data when results come in. End to end.

The second is a ground-up redesign of how study creation works inside Outset. A conversational flow that generates a guide and screener together, under 30 seconds, with quality checks surfaced inline before anything goes live.

Two different surfaces. The same idea.

What this actually looks like

Say an insights team gets a brief. A stakeholder wants to pressure-test an assumption before a key decision. The researcher is already working in Claude or Notion. With the MCP connected, they take the brief, create a study in Outset, configure recruitment, and launch, without switching tools or losing momentum. By the time they're done drafting the debrief outline, the first interviews are coming in.

The research team moves faster. The stakeholder gets answers before the window closes.

That's not hypothetical. It's a workflow pattern we've seen teams start building toward as soon as the read side of research became available in agentic tools. The missing piece was always the write side.

What makes ours different

A number of research platforms released MCP servers this spring. All of them are read-only. Connect them to your agent, and the agent can retrieve what your research found. That's useful. It's also only half the job.

The other half is creation.

Building a read layer is integration work. Building a write layer requires study creation infrastructure that already exists, is AI-native, and is reliable enough to hand to an external agent. That infrastructure took years to build inside Outset. The MCP is what makes it available to everything else.

The specific distinction: an agent connected to those aforementioned platforms can only ask what your research found. An agent connected to Outset can also create and launch the research that it pulls insights from.

What changed inside the product

The MCP is one side of this. The study creation redesign is the other, and it matters for the same reason.

A truly agentic flow in Outset is a conversation. Describe your goal or pick a use case from the quick-start menu. The agent asks a few clarifying questions, then generates a guide and screener together in under 30 seconds. Quality checks surface inline — flagging leading questions, probing gaps, question ordering issues — so you can adjust before anyone sees the study. Editing is collaborative throughout.

The bigger shift

We think about it this way: the Outset Agent works inside the product and outside it.

Inside: a researcher opens Outset and builds a study through a conversational AI that understands research methodology, catches quality issues inline, and generates the guide and screener together.

Outside: an agent in Claude, Cursor, ChatGPT, Notion, or Microsoft Copilot creates and launches that same study programmatically, without a researcher touching the interface at all. And helps them analyze it.

No other research platform can do both. That's what it means to be a truly agentic research platform.

Research should be where decisions get made. These two releases are the next step toward that.

Book time with the team to see Outset Agent in action →