by Sift Science, Inc.
AI fraud decisioning for payments, logins and sign-ups, scoring each event in under 150 ms against a network of 1 trillion events a year.
Sift suits fraud teams at marketplaces, delivery and subscription apps that fight the same people across fake accounts, takeovers and stolen cards; more than 700 customers use it. Small stores on Stripe and buyers who need published prices should look elsewhere first.
A user-level identity across sign-up, login and payment, informed by activity on the rest of Sift's network.
Its value grows with the network: a user seen across 34,000 sites and apps carries history a single merchant cannot build, which is lost if you leave.
Main trade-off
Quote-only pricing and an enterprise sales process put it out of reach of small teams.
Sift does not publish prices. Payment Protection, Account Defense and the AI agents are sold through a demo and a custom quote; the MCP server is included at no additional cost, per Sift's AI page.
Sift is an AI fraud decisioning platform that gives every sign-up, login and order a Sift Score and turns it into an accept, review or block decision. It learns from a shared network of more than 34,000 sites and apps, so a user flagged elsewhere arrives already known.
Sift Score and Workflows
Each sign-up, login or order is scored and routed to accept, review or block by rules you configure, with no manual step in the path.
Payment Protection
Decides on orders and payments to cut chargebacks while approving more good customers, with a Sift Score API for teams building their own flow.
Account Defense
Behavioural signals and a device fingerprint that persists across sessions catch account takeover at and after login, with step-up by email or SMS.
Review Agent
Surfaces the risk and trust signals in a case, explains them in plain language and suggests the decision to an analyst.
Fraud Attack Defense Suite
Agents that alert on attacks such as a BIN tested with small charges or a burst of fake accounts, then group the accounts behind it.
Sift MCP Server
Lets an AI assistant query a user's summary, labels and decisions in Sift from the tools a team already uses.
| Tool | Price from | Free plan | Runs on | Rating |
|---|---|---|---|---|
| Quote | No | Web | 3.8 | |
| $699/mo | Trial | Web | 3.8 | |
| $10/mo | Yes | Web | 3.9 | |
| -- | Trial | Web · iOS · Android | 4.1 |
Proprietary
Sift does not list prices on its website; every plan is a custom quote after a demo. Its AI page says the MCP server comes at no additional cost, while premium services are sold on top of the platform.
No free plan is offered. The way in is a demo request; Sift also runs a free quiz that scores your fraud operations in under two minutes.
It is a risk score Sift gives each user event, such as an account creation, login or order, based on that user's behaviour and the wider network's history. Your Workflows then map score ranges to accept, review or block, and Sift says that happens in under 150 milliseconds.
It judges each session rather than the whole account. When one login looks wrong, Sift can challenge just that session with an email or SMS code, so the real owner keeps using the account while the suspect session is stopped.
Yes. The Sift MCP Server lets an assistant pull a user's summary, labels and past decisions to triage a possible false positive inside the tools a team already uses, without a custom integration.
Sift is a fraud decisioning platform from Sift Science, Inc., led by chief executive Marc Friend. It sells two products on one score: Payment Protection, which decides whether to approve an order or payment, and Account Defense, which watches logins and sessions for account takeover. Sift says more than 700 customers use it, including DoorDash, Yelp, Poshmark, Patreon and ChowNow.
The mechanism is a shared network. Sift says its models see over 1 trillion events a year from more than 34,000 sites and apps, and its Global Profile links a user to activity seen elsewhere on that network. Every event (an account created, a login, an order) gets a Sift Score, and Workflows turn scores into accept, review or block decisions. Login protection adds behavioural signals and a durable device fingerprint, and acts on the single session, so one strange login triggers an email or SMS step-up instead of locking the whole account.
It fits fraud and trust teams at marketplaces, food delivery, travel, online gambling and subscription software, where the same bad actor shows up as a fake account, then a stolen card, then a hijacked login. Sift now also sells AI agents for those teams: a Review Agent that suggests a decision and explains it in plain language, a Risk Activity Agent that summarises a user's history, attack detection and investigation agents, a Workflow Optimization Agent, and an MCP server that lets an outside AI assistant query Sift.
Sift is more than a small Stripe store needs, where Stripe Radar is already switched on. Teams that want transparent, published per-check pricing and readable rules may prefer SEON, and merchants who want chargebacks paid for rather than predicted look at guarantee vendors such as Signifyd. See more on the AI security tools list.
Product facts on this page (pricing, plans, integrations, compliance) are read from Sift Science, Inc.'s own pages and documentation. The verdict is Hokai's reading of them, not the vendor's claim.