PickFu for Agents

Consumer research,
built for autonomous workflows

Run surveys from the terminal with the CLI or connect any AI agent via the MCP server. Structured JSON output, token-efficient polling, and zero-install execution.

Install
curl -fsSL https://agents.pickfu.com/install.sh | bash
Documentation pickfu.com
pickfu
$ pickfu survey list --status completed --json
[
{
"id": "s_8kf2m",
"name": "Logo preference test",
"status": "completed",
"responses": 50,
"winner": "Option A"
}
]
$ ▌

Works with

Claude
Cursor
VS Code
Windsurf
OpenAI
+
Any MCP client

Install

copied! curl -fsSL https://agents.pickfu.com/install.sh | bash

Downloads the latest binary. No Node.js required. Click to copy.

Built for agents, usable by humans

Every command supports --json for structured output. Errors tell you what to do next. Long-running operations block at the process level so agents don't burn tokens polling.

Create surveys programmatically

Define surveys as JSON, publish from the terminal. No browser needed.

$ pickfu survey create --from-file survey.json
$ pickfu survey publish s_8kf2m
Watch completions — no token burn

Process-level blocking. The agent's LLM isn't called while the CLI polls.

$ pickfu survey watch s_8kf2m --json
# blocks until complete — zero tokens consumed
MCP server for Claude & AI tools

Model Context Protocol integration. Works with Claude Code, Cursor, and any MCP client.

# In your MCP config:
"pickfu": { "url": "https://mcp.pickfu.com/mcp" }
API key auth — no browser

Store keys via stdin. Never appears in shell history.

$ echo "sk_..." | pickfu auth set-key
$ pickfu survey list # authenticated

Get started

curl -fsSL https://agents.pickfu.com/install.sh | bash
Read the docs Create a PickFu account