Highlights
Introducing Incode Know Your Agent Introducing TrustGraph Introducing GovFaceMatch 3 years in a row named a Leader
01/04
01/04
Back to blog
Incode Know Your Agent logo

Introducing Incode Know Your Agent, the end-to-end trust layer for AI agents, built on human identity

Today, Incode launched Incode Know Your Agent, which ties every AI agent to a verified person and the mandate that person sets. With Incode Know Your Agent, businesses can get access, control, and visibility over every AI agent that acts on behalf of their customers.

AI agents already act for people in banking, commerce, and everyday services. A fraudster's agent can arrive with the same valid credentials as a customer's, and at login a business can't tell the two apart. Incode Know Your Agent closes that gap by putting one verified person behind every agent.

Incode Know Your Agent detects AI agents, verifies the person behind each one, enforces the mandate that person sets, and records every action. Deepsight, GovFaceMatch, and TrustGraph strengthen each check with deepfake detection, one-to-one government matching, and privacy-preserving network detection.

Until now, institutions have had to choose: let every AI agent through the door and invite agentic fraud, or block every AI agent and lose the customer. With Incode Know Your Agent, that choice ends.

AI agents now bank, shop, and pay on behalf of customers

AI agents are creating a new kind of autonomous, fast-moving economy. Customers and attackers alike send agents into that economy, often to the same businesses.

Customers are already sending AI agents to act for them

Nearly seven in 10 Americans (69%) say they would let a bank's AI assistant complete at least one financial task for them, according to TD Bank. The commercial stakes are large: McKinsey estimates that AI agents could orchestrate $900 billion to $1 trillion in US retail sales by 2030.

The agents customers send log in, pay, and book with the customer's own credentials. For a business, each one represents a customer who chose to delegate.

How attackers use AI agents to run fraud end to end

Fraudsters send agents too. An attacker's agent can run a whole attack with no person involved. It steals an identity, builds the evidence, makes an attempt, reads the response, adjusts, and tries again.

On Incode's platform, AI-assisted fraud attempts rose from 2% in Q1 2023 to 32% in Q1 2026, a 16-fold increase in three years. Incode projects that more than 60% of fraud attempts will be AI-assisted by the end of 2026. Across industries, 81% of companies say they have already faced attempted or successful agentic fraud,* according to KPMG. Incode's Agentic Fraud Report tracks how these attacks are evolving.

*Agentic fraud: fraud that is AI-generated, AI-assisted, or agent-operated.

Why businesses can't tell legitimate AI agents from fraudulent ones

At login, a customer's agent and a fraudster's agent look identical. Both present a valid credential, a clean device, a residential IP address, and the correct one-time code.

That leaves businesses with a hard choice. If they block every agent, they can lose their customers. If they let every agent in, fraud can come in with them.

This fall, some of the largest US companies have chosen to block agents. Every blocked agent can be a customer turned away, free to take their business to a competitor that lets the agent in.

Introducing Incode Know Your Agent

Incode Know Your Agent lets businesses welcome AI agents under a mandate each customer sets. In one end-to-end flow, it shows a business who is behind each AI agent, what it may do, and what it has done.

It detects the agents in a business's traffic, including those that never announce themselves. It then verifies the person behind each agent with a live face check against the government's own record and binds the agent to that person. That person sets the agent's mandate: which tasks it can do, how much it can spend, and for how long. Every action is checked against that mandate and recorded under the person who authorized it. An agent with no verified person behind it is stopped before it acts.

Incode Know Your Agent works across payment rails and AI agent providers. Because Incode doesn't operate any of them, every party can trust the record it keeps. Verifying the human behind each agent is the core of Know Your Agent (KYA).

For businesses, Incode Know Your Agent delivers three outcomes:

  1. Meet customers where they already are. Admit the agents your customers send to pay bills and manage accounts, with the same verification you already run on the person.
  2. Catch agentic fraud at the door. Block any agent with no verified person behind it before it acts and before a loss reaches your books.
  3. Know who is accountable. Tie every agent action to the person who authorized it and the mandate they set, so a questioned action has a name.

How Incode Know Your Agent works

A single session carries each agent from first detection to a permanent record, in this order:

  1. Agent Detection finds the agent. Browser behavior and automation signals run on every session, and a live-presence check runs when those signals point to an agent. Once Incode Know Your Agent spots an agent, the business chooses whether to register it, step up to the person, or block it.
  2. Agentic Identity links the agent to its person and mandate. The customer verifies once, live, and gives consent. Each agent then carries that verified identity and a mandate covering scope, amount, and duration. One person can run many agents, and each agent answers to one person.
  3. Agent Access Management decides what the agent can do. Before any action runs, it checks four things: a registered agent, a valid mandate, a verified person, and an action that falls within the mandate. When an action goes beyond the mandate, the person approves it live with a face check on their phone.
  4. Agent Monitoring blocks compromised or misaligned agents. It reads patterns across sessions and organizations, so an agent blocked at one organization arrives at the next already flagged.
  5. Agent Audit records every action. A tamper-evident trail captures the action, the person, the mandate, and the proof behind it, visible to both the business and the customer. The business also sets which agents may do what, action by action.

An example mandate in action

Consider a customer who sends an agent to handle everyday payments.

The customer sets a mandate that lets the agent pay up to $500 per transaction on its own, with a ceiling of $1,200 a month. When the agent tries to make a payment above $500, Agent Access Management holds the payment and sends a face check to the customer's phone. Once the customer approves, the payment goes through, and Agent Audit records the action, the person who approved it, and the proof of that approval.

Built on Deepsight, GovFaceMatch, and TrustGraph

Incode already answers three questions about every person it verifies, and Incode Know Your Agent adds a fourth for every AI agent: who answers for it.

  • Are you real? Deepsight checks for a live person and catches deepfakes and injected camera feeds.
  • Are you who you claim to be? GovFaceMatch matches a live face against the photo on the person's state DMV record.
  • Are you a good actor? TrustGraph reads what an identity has done before, across more than 800 organizations.
  • Who answers for you? Agentic Identity binds the agent to the verified person behind it and the mandate they set.

That fourth answer rests on the person checks above. An agent can hold a credential, a device, and even a generated face, but only a live person can pass a liveness check and match the photo the DMV holds. GovFaceMatch produces 1,000 times fewer false acceptances than Social Security number checks, and in testing at Purdue University, Incode Deepsight held false acceptance rates on deepfakes to 0.2%, compared with 13% for the next vendor.

Agent protocols such as the Personal Agent Protocol (PAP) and the Model Context Protocol (MCP), the standard that connects AI agents to tools and data, carry the agent's credentials. Incode adds the verified person those protocols leave open.

Why organizations are building agentic trust with Incode

Incode Know Your Agent lands on an integration that many of the largest institutions already run.

More than 800 organizations trust Incode, including nine of the top 10 US banks, eight of the top nine US telcos, and four of the top five global marketplaces. Banks, marketplaces, telecoms, and government programs are among the organizations building agentic trust with Incode, adding the agent layer to the verification they already run on the person.

Get started with Incode Know Your Agent

Incode Know Your Agent is available today. Businesses integrate it with a web SDK on every page and a backend API for policy, and the developer docs are live.

Ready to see how Incode Know Your Agent puts a verified person behind every AI agent that reaches your business? Request a demo today.


Ready to see it in action?

Request a demo