How do I connect my agent to Fiuto?

Fiuto runs as a remote MCP server. From your editor, you can ask a connected agent to build studies, launch them, pull results, and wire tracking into a prototype.

This page covers what the MCP gives your agent, what you can ask it to do, and how to connect from any MCP-ready tool. For step-by-step setup in a specific tool, jump to its page below.

What the Fiuto MCP gives your agent

Once connected, your agent can work with Fiuto through the capabilities available to it:

• Browse your projects, studies, and the block catalog.
• Turn a research goal into a starting plan, with a recommended block sequence and example configs to edit before building.
• Draft a plan, then build and edit the blocks. Group studies into projects.
• Preview a study before launch.
• Launch a study and get back the share link and a results link.
• Pause, stop, or archive a live run.
• Check a launched run’s progress when your agent supports it: how many respondents have started, how many finished, and whether the run is live, paused, or complete.
• Read results from a completed run.
• Read your saved Analyses, including which studies each one draws on.
• Attach and verify behaviour tracking on a prototype your agent can edit. See Wiring Fiuto tracking into a prototype.

The available actions depend on the Fiuto tools exposed to your agent and your access to the study. Start in the agent and continue in the web app, or the other way around.

What you can ask your agent to do

You drive it in plain English. The agent translates your request into the right Fiuto actions, so you don’t need to know what the underlying tools are called.

A typical end-to-end session

1. “Plan a five-question study to find out whether our new sign-up flow is clear, then build it.”
2. “Launch the study and give me the share link.”
3. “Pull the results and tell me what people struggled with.”

More things to try

1. “Build a preference test between these two pricing pages, and ask for a one-line reason after each pick.”
2. “Show me my draft studies.”
3. “Add a screening question so only designers can take this study.”
4. “Pause the onboarding test, I need to fix a typo first.”

Connect from any MCP-ready tool

Any tool that supports remote MCP servers can connect to Fiuto, including ones not listed here. The shape is always the same:

1. Add the server. Point your tool at the Fiuto MCP server: https://mcp.fiuto.ai/mcp-server.
2. Sign in. The first time you connect, your tool opens a browser to sign in to your Fiuto account. There are no tokens or API keys to paste.
3. Verify. Ask your agent to list your studies. If the connection is live, it returns them, or an empty list if you haven’t made any yet.

You’ll need a Fiuto account first. Sign up at app.fiuto.ai.

If you use Claude on the web at claude.ai, Fiuto connects as a custom connector under Settings, then Connectors. This needs a paid Claude plan. Paste the same server URL, sign in, and toggle the connector on in a chat. The three steps above still apply.

Set up your specific tool

Step-by-step setup for the tools we see most. Each guide covers the connection, a quick check that it worked, and the snags particular to that tool.

Claude Code

Cursor

Codex CLI

Lovable

v0

Replit

Plans and usage

Connecting an agent does not add a separate allowance. Fiuto uses the plan and usage shared by your account, whether you work in the web app or through an agent. Open Settings → Plan & usage to see your current limits. If an action is unavailable, Fiuto explains what is needed and points you to the next step.

Tools your agent can call

Fiuto exposes its whole study loop as MCP tools. You don’t need to know their names to ask for something in plain English, but here’s the reference if you want it:

• Study and plan authoring:
  • get_fiuto_model, list_studies, get_study, list_projects, create_project, move_to_project, get_user_context
  • scaffold_plan, save_plan, build_from_plan, create_study, update_study, delete_study, bulk_delete_studies
  • list_blocks, add_block, update_block_config, delete_block, move_block, reorder_blocks, bulk_add_blocks, bulk_update_blocks, get_preview_url
• Launch and run control: launch_study, pause_spawn, stop_spawn, archive_spawn.
• Results and analysis: get_study_results, export_study_results, list_analyses, get_analysis.
• Prototype tracking: attach_tracker, verify_tracker, get_tracker_events, wrap_url, check_prototype_accessibility.
• Agent-run debugging: list_traces, show_trace, for inspecting a past agent run against your studies.

delete_study and bulk_delete_studies sit behind a dedicated delete scope that most connections don’t request, so those two may be missing from your agent’s own list even though they’re in the reference above.

Ask your agent to list its tools if you want to see exactly what your client has exposed.

Next steps

Author a study from your agent What to ask for to draft a study end to end, how to iterate, and the cleanup worth doing before launch.

Pull study results into your agent Read and analyse your study's responses from your agent. What landed, what's missing, what to ask next.

Wire Fiuto tracking into a prototype Capture respondent behaviour inside a prototype your agent built, and tie it to a study.