MoltPe

MoltPe Agent Payments

Payment infrastructure for AI agents — reference implementation with x402, MPP, and fiat payment paths.

Disclaimer: This is a reference implementation with mock data. For production, implement real PaymentProvider adapters. MoltPe's production server is at moltpe.com/mcp.

What This Is

This repository contains a reference implementation for handling payments in AI agent systems. It supports three payment paths: stablecoin micropayments (x402), session-based streaming (MPP), and traditional fiat rails. The goal is to show how agents can pay, get paid, and manage money across different protocols.

Quick Start

# MCP Server — payment tools for AI agents
cd mcp-server && npm install && npm start
# Server runs on port 3000 with 11 MCP tools
# Connect Claude Desktop with config from examples/claude-desktop-config.json

# x402 Client — stablecoin micropayments
cd x402-client && npm install && npm run build
# Use in your project:
# import { pay } from '@moltpe/x402-client'

See mcp-server/README.md and x402-client/README.md for full usage.

Demo

Start the server and open http://localhost:3000 to explore all 11 payment tools interactively.

The demo includes:

• Tool explorer with editable inputs for every MCP tool
• Live balance tracking across stablecoin, session, and fiat providers
• MPP session budget visualizer with real-time progress
• Invoice lifecycle: create → send → track → collect
• Transaction timeline of all actions

Run with mock data (default) or connect to MoltPe's live MCP server.

See mcp-server/DEMO.md for a step-by-step walkthrough.

Live Mode

cd mcp-server
PROVIDER_MODE=live MOLTPE_AGENT_TOKEN=swai_... npm start

Connects to moltpe.com/mcp for real on-chain transactions. See mcp-server/README.md for setup.

What's Inside

Directory	Description
research/	Market research, protocol analysis, competitive landscape
mcp-server/	MCP payment server with 11 tools, 3 mock providers (x402, MPP, fiat), 53 tests
x402-client/	TypeScript x402 client SDK with facilitator adapters and interceptors, 34 tests
playbook/	Build playbook: process, templates, results from building this
docs/	Architecture docs, protocol landscape, market analysis

Tri-Rail Architecture

This implementation supports three parallel payment paths:

• Path A — Stablecoin (x402): Per-request micropayments settled on-chain in USDC
• Path B — Session-based (MPP): Authorize a budget, stream payments within a session
• Path C — Fiat: Traditional card authorization, capture, and settlement

Each path serves different use cases. A collections layer sits above all three to handle agent receivables.

flowchart TD
    Agent[AI Agent] --> Collections[Collections Layer<br/>Invoicing · Billing · Receivables]

    Collections --> A[Stablecoin Path<br/>x402]
    Collections --> B[Session Path<br/>MPP]
    Collections --> C[Fiat Path<br/>Card / Bank]

    A --> A1[USDC Micropayments]
    A1 --> A2[On-chain Settlement<br/>Polygon · Base]

    B --> B1[Budget Sessions]
    B1 --> B2[Streaming Settlement<br/>Tempo Network]

    C --> C1[Card Authorization]
    C1 --> C2[Gateway Settlement<br/>Razorpay · Stripe]

    classDef agent fill:#0ea5e9,stroke:#0284c7,color:#fff
    classDef collections fill:#f59e0b,stroke:#d97706,color:#fff
    classDef x402 fill:#22d3ee,stroke:#06b6d4,color:#000
    classDef mpp fill:#a78bfa,stroke:#8b5cf6,color:#000
    classDef fiat fill:#34d399,stroke:#10b981,color:#000

    class Agent agent
    class Collections collections
    class A,A1,A2 x402
    class B,B1,B2 mpp
    class C,C1,C2 fiat

See docs/tri-rail-architecture.md for the full breakdown.

Research Highlights

• Verified agent payment volume is $1.6M/month — not the $43M self-reported (48-86% is artificial). See research/data/verified-volumes.md.
• Agent collections/receivables is a near-total market gap — only PayPal MCP exists, ecosystem-locked. See docs/collections-gap.md.
• Four protocols (x402, MPP, ACP, AP2) are converging into layers, not competing head-to-head. See docs/protocol-landscape.md.

Quality & Security

Metric	Value
Tests	87 (53 mcp-server + 34 x402-client)
Pass rate	100%
Statement coverage	86.2% / 95.9%
Runtime dependencies	0 (both packages)
npm audit	0 vulnerabilities
Secret scan	0 findings

See QUALITY.md for full coverage breakdowns and scan results. See SECURITY.md for vulnerability reporting and security posture. See docs/production-security-checklist.md for the 64-item production deployment checklist.

Integration Examples

Looking to integrate MoltPe Agent Payments with your AI framework? These issues track planned integrations with example code:

• LangChain tool wrapper — Use MCP payment tools as LangChain tools
• AutoGen agent — Payment-capable AutoGen agent with budget guardrails
• Base Sepolia testnet — Real on-chain x402 payments on testnet
• Webhook notifications — Event-driven payment notifications
• Security hardening docs — Threat model and hardening guide
• Eliza plugin — Payment plugin for the Eliza agent framework

Contributions welcome — pick an issue and open a PR.

Deep Dives

Long-form guides covering the MoltPe stack, AI agent payment patterns, and India-specific integration:

For developers evaluating MoltPe

• Why developers choose MoltPe for AI agent payments — five honest reasons, including the tradeoffs
• Integrate MoltPe in 5 minutes: developer quickstart — REST, MCP, x402, Python — all four integration paths with real code
• MoltPe vs building your own AI payment stack — the build-vs-buy analysis with real time estimates
• MoltPe developer experience: a deep dive — error messages, docs, testnet, rate limits
• The MCP server for AI agent payments, explained — Claude Desktop / Cursor / Windsurf config + tool surface

For Indian developers and AI startups

• AI Agent Payments in India: The Complete Infrastructure Guide (2026) — pillar guide for India
• The Cost of AI Agent Payments in India: 2026 Benchmark — 4-scenario cost comparison, 15-28x savings over PayPal
• USDC payments for Indian developers — USDC vs Razorpay vs Stripe India vs PayPal
• Freelance AI developer payments in India — the forex-avoidance playbook
• The complete payments stack for Indian AI startups — 3-layer stack: UPI + MoltPe + agent-to-agent
• x402 protocol for Indian developers — API monetization in USDC

Protocols and fundamentals

• What are AI agent payments?
• The x402 protocol: complete guide
• Machine Payments Protocol (MPP) guide
• AI agent spending policies explained

Short answers (Q&A format)

• What is an AI agent wallet?
• How do AI agents make payments?
• What is an AI agent spending policy?
• How to monetize an API in USDC

Also: FAQ · Glossary · Use cases · Developer docs

License

• Code (everything outside research/, playbook/, docs/): Apache License 2.0
• Content (research/, playbook/, docs/): Creative Commons Attribution-ShareAlike 4.0

See NOTICE for attribution details.

Citation

@software{moltpe_agent_payments,
  author = {Gupta, Umang},
  title = {MoltPe Agent Payments},
  year = {2026},
  url = {https://github.com/umangbuilds/moltpe-agent-payments},
  license = {Apache-2.0}
}

Author

Created by Umang Gupta. MoltPe builds payment infrastructure for AI agents.