AgentPass -- Secure MCP for Regulated Finance. MCPS + Agent Identity for the Agent Economy. IETF Draft
AgentPass · Agent payments

Any agent. Any rail.
One pass.

Rapid payment integration supporting any payment flow or standard. AgentPass sits between your agent and the payment rail -- signing identity, scoring trust, screening sanctions, and shaping the request into whatever the rail wants. Works with x402, AP2, ACP, MPP, Mastercard Agent Pay, Visa TAP, Stripe, and L402.

See 60-second demo Integrate in 10 lines
McKinsey projects $3T--$5T of global agentic commerce value by 2030 and up to $1T of US retail revenue directed by agents. Someone has to verify every one of those payments.
Linux Foundation x402 ready AP2 Mandate compatible Visa TAP signature compatible Mastercard sandbox integrated OWASP MCP Top 10 contributor IETF draft-sharif-* Cited in academic research (arXiv)

The architecture of autonomous agent trust

One trust layer. Any transport. Any payment rail.

The architecture of autonomous agent trust: interaction and settlement layer (MCP, WebMCP, A2A, REST, CLI audited by xat), unified payment gateways (Stripe, L402, x402, ACP, MPP, TAP, Agent Pay), validation and trust layer (x-agent-trust ECDSA P-256, AP2 Mandates, Visa TAP, ERC-8004), and AgentPass Trust Scores (L0-L4: AML, sanctions, velocity).
Secure Agent Identity x-agent-trust (ECDSA P-256, AP2 Mandates, Visa TAP, ERC-8004) Audited Transport MCP, WebMCP, A2A, REST, CLI -- all signed by xat Unified Payment Gateways Stripe, L402, x402, ACP, MPP, TAP, Agent Pay AgentPass Trust Scores L0--L4, graded on AML, sanctions, transaction velocity

No token tax per payment

MCP-native payment servers burn the context window on every transaction. AgentPass doesn't.

30-tool payment MCP
Expensive
3,705tokens schema floor
  • 30 tools: create_quote, confirm_payment_intent, capture, refund, dispute, customer CRUD, transfer, payout, subscription, webhook_verify, sanctions, KYC, velocity, risk, flag review
  • Full JSON schema re-injected on every hop
  • Real payment flows are multi-hop (quote → confirm → capture ≈ 11,115 tokens)
  • Tool-selection accuracy collapses as menu grows
$333 / month at 1,000 payments/day (Sonnet)1
AgentPass
16× smaller schema · ~48× effective
230tokens schema floor
  • One tool: agentpass_pay(rail, to, amount, currency)
  • Agent signs once with x-agent-trust (ECDSA, zero tokens)
  • Trust + sanctions + AML + rail settlement all run in deterministic code
  • Single call -- LLM never re-enters the loop
$21 / month at the same volume (saves $313)
The principle
AI belongs in the decision layer -- "should I authorise this payment?" After the agent signs, the payment rail is execution layer: deterministic code, zero tokens, full audit trail from the signature. This is Armin Ronacher's "code is the tool" pattern applied to money.
Numbers measured with tiktoken cl100k_base against a realistic 30-tool payment MCP and our 1-tool agentpass_pay. On Claude Opus ($15/MTok input) all figures x5: $1,667 / month vs $104 / month. Reproducible: github.com/cybersecai-uk/agentpass-demo-mcp/tree/main/benchmark.
1 3,705 tokens x 1,000 req/day x 30 days ÷ 1,000,000 x $3 ≈ $333. Agent payment flows that span multiple tool hops (quote -> confirm -> capture) multiply the 30-tool figure by the number of hops, giving an effective ~48x advantage at the payment-flow level.

Every credible agent payment rail

One adapter. One interface. One key pair.

Click-to-pay, for agents

Pick a rail. Pick an agent. Watch AgentPass gate the payment in real time with live sanctions and Mastercard risk scoring.

Integrate AgentPass into your tech stack

AgentPass is not just for AI chatbots. Any system that acts autonomously needs cryptographic identity, trust gating and an audit trail. Three lines of code. Any language. Any device.

$ pip install agentpass-sdk

from agentpass import AgentPass
ap = AgentPass(endpoint, api_key)

# That is all it takes to add cryptographic identity
# and trust enforcement to any autonomous system.

Where AgentPass drops in

Robotics
Warehouse robots, humanoid agents, factory automation. Every motor command is trust-gated. Forklift operation requires L4 human approval. Every movement signed and logged.
IoT & Smart Home
Smart locks, thermostats, cameras. Adjusting temperature is L1. Unlocking the front door is L4 with hardware human approval. No autonomous device should act without verifiable identity.
Payment Systems
POS terminals, mobile wallets, agent commerce. Every transaction is MCPS-signed with sanctions screening. Large payments require human approval. Full audit trail for PCI and regulatory compliance.
Autonomous Vehicles
Lane changes at L2. Emergency braking at L2. Autopilot override at L4 with human approval. Every driving decision is cryptographically signed and forensically auditable.
Industrial Automation
SCADA systems, PLCs, factory controllers. Production line shutdown requires L4. Maintenance scheduling at L2. Every action signed into a tamper-evident ledger.
Medical Devices
Drug dispensers, surgical robots, patient monitoring. Dosage changes require L3. Surgical actions require L4 with human approval. Audit trail meets regulatory evidence requirements.
Enterprise Software
Any application with embedded AI agents. Database access, code deployment, data export, API integrations. Same trust model, same audit trail, same kill switches.

Available SDKs

Python
pip install agentpass-sdk
Node.js
npm install @agentpass/sdk
C / C++
agentpass.h (embedded)
REST API
any language, over HTTPS

Every agent gets a certificate. Every action gets a signature. Every decision is trust-gated from L0 to L4. Every critical action requires human approval. Every event lands on a tamper-evident audit trail.

If it acts autonomously, it needs AgentPass.