Developers

Build with Pickrate

Two products, one funnel. Pick Rate measures whether AI agents choose your tool; Agent Attribution measures whether that turned into revenue. Both are built for machines to read and call.

New here? Start with setup

Get your agent traffic into Pickrate, then ask Claude or ChatGPT how much traffic and how many signups AI agents are driving.

Set up Pickrate →

Agent Attribution

Tell Pickrate when an agent drove a visitor who later converted. Three events (touch → identify → convert), a server SDK, a browser helper, signed webhooks, and a CSV/JSON export.

Integration guide →

Pick Rate API

Pull any tool's Pick Rate, rank a category, or run your own evals. A public read API, an authenticated eval API, and an MCP server agents can call directly.

API reference →

Install the collector

Get your agent traffic in: a Vercel log drain or a Cloudflare Worker (both no-code), a one-line Next.js middleware, or a pasted access log. Agents don't run JavaScript, so a browser tag can't see any of this.

Install guide →

Quick start

Send your first attributed conversion in three steps:

# 1. Get a secret key in Settings → Agent Attribution keys
# 2. Install the SDK
npm i @pickrate/attribution

# 3. Record the money event
import { createClient } from "@pickrate/attribution";

const pr = createClient({ secretKey: process.env.PICKRATE_SECRET_KEY });
await pr.convert({ email: "user@acme.com", kind: "signup" });

Full walkthrough — the visitor → identity bridge, the browser helper, getting your data back out — in the attribution guide.

Machine-readable surfaces

Everything an agent needs to discover and call Pickrate without a human:

Keys & access

Attribution keys (pk_ / sk_) and Pick Rate API keys are both minted in Settings. The public read API and the MCP server need no key. See how the number is made for what the data means.