The trust handshake for the agent economy

Trust any agent before it transacts.

AgentFide verifies any agent, scores its reputation, and backs the deal with recourse, all in a single call your agent makes before money moves.

Start building, freeRead the docs
awaitagentfide.assess({ agent, action })// → allow

Self serve from first call to first transaction. No demo required.

POST /v1/assess
~120ms
Your agent
AgentFide
Counterparty
  • Identityverified
  • Authorityin scope
  • Reputationgold
  • Recoursecovered
decisionallow

Agents can transact now. Trusting them is the hard part.

Your agent can pay, hire, and buy in milliseconds. Every counterparty is stuck on the same question: should it trust the agent on the other side. Is it who it claims to be. Is it authorized to spend. Does it have a track record. And what happens if it goes wrong. Today there is no clean answer, so real money waits on human review.

How it works

One call. Four checks. A clear decision.

1

Identity

We verify the agent's signed credential and the operator behind it.

You know who you are dealing with.

2

Authority

We confirm the agent is acting within a valid, scoped mandate.

You know it is allowed to do this.

3

Reputation

We return a transparent, explainable score built from its real track record.

You know whether it can be trusted.

4

Recourse

We back the transaction with escrow and the ability to undo a bad one.

You know you are covered.

The single call

Reachable in one tool call.

AgentFide ships as an MCP server and drops into the major agent frameworks. Your agent calls assess before it transacts, gets back a clear allow, review, or deny, and a credential it can show the other side. If you already use the Stripe Agent Toolkit, AgentFide installs right alongside it.

  • MCP server
  • OpenAI Agents SDK
  • Vercel AI SDK
  • LangChain
  • CrewAI
  • Stripe Agent Toolkit
assess.ts
// Installs alongside your payment tools — same setup.
import { agentfideTool } from "@agentfide/sdk/openai";

const agent = new Agent({
  tools: [stripeTool, agentfideTool],  // payments + trust
});
Add the trust check in the same setup as your payment tools.

Recourse

When an agent gets it wrong, undo it.

A score tells you the odds. Recourse lets you act when the odds break. AgentFide holds funds in escrow and can roll back a bad transaction, so a single mistake is recoverable, not final.

↺Watch a rogue transaction get caught and undone, live.Run the demo →

Live demo

Catch a rogue agent. Undo a bad deal.

The same logic that runs in production, on example agents. Clear a bona fide agent, catch a rogue one before money moves, then undo a bad outcome with Agent Rewind.

A bona fide agent transacts

PayBot, operated by Acme

Buy $42 of API credits from a verified seller

  • Identity. Signed A2A card verified. Operator: Acme.
  • Authority. Within the $500 treasury mandate.
  • Reputation. platinum, score 980 from a clean record.
  • Recourse. Covered. Coverage fee $0.42.
ALLOWAll four checks clear. The deal proceeds with escrow backing it.
Escrow holds $42Credits deliveredReleased to seller

Cleared and settled. The seller is paid, the buyer got what it paid for.

Act 1 of 3

Built on the standards agents already speak.

Neutral by design, owned by no rail or marketplace, so everyone can feed it and rely on it.

A2A signed agent cardsModel Context ProtocolAP2 mandatesStripe Agent Toolkit

The standards solve one layer each, and leave the rest open. Read The State of Agent Trust 2026 for the evidence.

Pricing

One plan. $49/month.

The whole assess trust API for one flat price. Enterprise recourse coverage and volume when you need them. We never charge you to keep a transaction private, and an agent can never buy a better score.

See pricingContact sales

Give your agents a reason to be trusted.

Start building, freeRead the docs

From zero to your first verified transaction in minutes.