Projects

The project is Faraday Pro's core abstraction. A project is a saved payload — the named set of context elements you want about a person — plus the ways you get that payload back out.

Everything downstream reads from it. A file append, a real-time lookup, an agent calling MCP, and a warehouse deployment all return the same payload, because they all read the same project. Change the project's selection and every one of them changes together.

Creating a project

Projects are how you organize your work. Each has its own set of context elements, so create a new one anytime you want a different payload — one project for scoring inbound leads, another for appending mailing addresses to a win-back list.

Your first project opens pre-loaded with a set of featured attributes, so there's something in the payload before you've chosen anything. Replace them as soon as you know what you actually want.

Projects are independent. Adding an attribute to one never affects another.

Building your payload

Use the Add elements button to open the element chooser. It has three tabs, and the three kinds bill at different rates.

Consumer profile attributes

The 1,400+ attributes in Faraday's catalog: demographics, financial indicators, property details, shopping behaviors, life stage, and more.

If you know what you want, search for it. If you don't, describe the problem instead — the suggester works semantically, so "datapoints around home ownership" surfaces the right fields without your knowing what they're called.

Identity

The person's own contact and address details, appended to the rows we match. Ten elements are on offer:

Identity is what you reach for when you hold one channel and need another — you have emails and want mailing addresses, or you have addresses and want to reach people by phone. The Faraday Person ID is a stable identifier for the matched person, useful as a join key across your own systems.

You are never charged for an identity element you supplied yourself. On a file, that's anything your identity mapping supplies; on a lookup, anything your request already contained. Only genuinely new identity is billed.

Custom predictions

Predictions are context that doesn't exist yet — something Faraday models rather than looks up. Two kinds:

• Propensity — the likelihood a person does a particular thing: converts, buys again, lapses. You define it by pointing at a group of people who already did it. Returns a probability, a percentile, and a score.

• Persona — a data-driven clustering of your people into named groups. Each matched person comes back with their assigned persona, and you can optionally choose which attributes the clustering should key on.

Both are built from segments: reusable groups of your own people, created by uploading a CSV or pulling from a connection. A propensity needs a segment of people who attained the outcome, and optionally a second segment of people who didn't, which sharpens the model. A persona set needs one segment to cluster.

Predictions live at the account level, not inside a single project. Once you've built one, you can attach it to any other project without rebuilding it.

Getting your context

Four ways out, all reading the same payload.

Real-time lookup by API

POST an identifier to your project's append endpoint and get the payload back immediately:

POST https://api.pro.faraday.ai/v1/projects/{project_id}/append
Authorization: Bearer YOUR_API_KEY
Content-Type: application/json

{ "email": "someone@example.com" }

To match a person, send at least one of these combinations:

• email
• phone + last_name
• address + city + state
• address + zip

The response carries the matched attributes, any identifiers you selected, any predictions on the project, and what the call cost — billed_credits alongside your remaining balance_credits, so you can meter spend without a second request.

This is the right surface for scoring a lead the moment it arrives, or enriching a record mid-workflow. It's also the simplest thing to point a CRM, CDP, or ESP webhook at.

Real-time lookup by MCP

The same lookup, addressable by an AI agent. Point an MCP client at:

https://api.pro.faraday.ai/v1/mcp

Connection is over OAuth — your client runs the consent flow on first use, so no API key goes into a config file. Beyond the lookup itself, the server exposes tools for listing and managing projects and for browsing the attribute catalog, so an agent can assemble its own payload and then use it.

This is what makes Faraday Pro addressable from Claude, ChatGPT, or anything else you're building with: the agent retrieves real context about a person on demand rather than guessing.

Append to a file

Upload a CSV, map your columns to identity fields, and download the file back with your payload appended. Files can be up to 5 GB, and there's no row cap.

You'll see the match rate and the exact credit cost before you download. Nothing is charged until you click download — an upload that matched worse than you hoped costs nothing to walk away from.

Your payload is snapshotted when you submit the mapping, so editing the project afterward won't change a file already in flight, and the delivered column headers stay stable.

Connect your warehouse

Point a project at a table in your data warehouse, database, or cloud bucket, and Faraday enriches those rows and writes the results to a new table.

Cost works differently here, and better: the rows are enriched into a measurement pass first, so you're told the exact price and asked to confirm before anything is delivered or charged. You can also route the output to a downloadable CSV instead of a table.

One difference worth knowing: your identity selection doesn't apply to warehouse deployments. Faraday writes those tables directly and emits the full identity set regardless of what you picked — so identity comes along, but you aren't billed for it.

How billing works

You're charged per filled value, not per row submitted and not per attribute selected. A person we can't match, or an attribute that comes back blank for someone, costs nothing.

So a project with 20 attributes and one propensity, run against a list where 8,000 of 10,000 rows match, bills on the values actually returned for those 8,000 — not on 10,000 × 21.

See pricing for credit rates and volume discounts.

Next steps

• About Faraday Pro — what Pro is, and how it compares to Faraday Enterprise.
• The data catalog — every attribute you can add to a payload.
• MCP — more on Faraday's agent-facing surface.