Ag402 primary sources — 2026-05-03

Project status in this pass

  • No canonical standalone Ag402 whitepaper or litepaper surfaced.
  • The clearest current source of truth is the main GitHub repository, its README, and the project llms.txt.
  • Ag402 is best understood as GitHub-first x402 payment middleware for both buyer-side agent autopay and seller-side API/MCP monetization.

Primary-source snapshots

1) Repository landing page / high-level positioning

  • URL: https://github.com/AetherCore-Dev/ag402
  • Key excerpt:
    • The repository landing page says Ag402 gives an AI agent a wallet and lets it pay for APIs automatically.
    • It explicitly describes the flow as 402 Payment Required → auto-pay USDC on Solana → retry → 200 OK.
    • It also highlights zero code changes, non-custodial operation, zero telemetry, support for Claude Code/Cursor/OpenClaw, and seller-side monetization through ag402 serve.

2) Main README / product architecture and security posture

  • URL: https://raw.githubusercontent.com/AetherCore-Dev/ag402/main/README.md
  • Key excerpt:
    • The README says Ag402 is “the payment layer for the Coinbase x402 protocol” and repeatedly frames the project around buyer-side, seller-side, and MCP-side zero-code integration.
    • It claims support for Python monkey-patching, TypeScript fetch wrappers, reverse-proxy paywalls, MCP auto-installation, Solana-USDC settlement, prepaid payment paths, and multi-framework compatibility.
    • It also emphasizes 775+ tests, 90%+ coverage, multiple security-review rounds, six-layer budget protection, seller-no-key mode, and zero telemetry.

3) llms.txt / package decomposition and AI-tooling posture

  • URL: https://raw.githubusercontent.com/AetherCore-Dev/ag402/main/llms.txt
  • Key excerpt:
    • The llms.txt file presents Ag402 as “Payment infrastructure for AI agents. x402 protocol + Solana USDC.”
    • It breaks the stack into four packages: open402, ag402-core, ag402-mcp, and ag402-client-mcp, which clarifies that the project spans protocol, buyer runtime, seller gateway, and MCP tooling.
    • It also documents explicit installation paths for Cursor, Claude Code, and OpenClaw, plus many CLI commands for wallet, history, prepaid credentials, health checks, and safety limits.

4) Buyer-side zero-code integration

  • URL: https://raw.githubusercontent.com/AetherCore-Dev/ag402/main/README.md
  • Key excerpt:
    • The README shows that Python users can run ag402 run -- python my_agent.py or add ag402_core.enable() and leave their agent logic untouched.
    • It says this works with LangChain, AutoGen, CrewAI, Semantic Kernel, and any framework using httpx or requests.
    • This is a strong clue that Ag402 is focused on invisible payment handling at the HTTP layer rather than a special-purpose app framework.

5) Seller-side and MCP-side monetization

  • URL: https://raw.githubusercontent.com/AetherCore-Dev/ag402/main/README.md
  • Key excerpt:
    • The README says sellers can run ag402 serve --target ... --price ... --address ... to wrap an API or MCP server with a paid proxy.
    • It explicitly claims zero code changes for both API sellers and MCP server developers.
    • This is important because it places Ag402 on both sides of the machine-payment handshake, not just on the buyer side.

6) Production examples

  • URL: https://raw.githubusercontent.com/AetherCore-Dev/ag402/main/README.md
  • Key excerpt:
    • The project highlights Token RugCheck as a live Solana-mainnet example where agents pay $0.02 USDC per audit.
    • It also points to a DeepBlue API example for low-cost BTC/ETH/SOL trading signals.
    • These examples support the project’s claim that it is already being used for real paid machine queries rather than only toy demos.

Working conclusion

Ag402 is best cataloged as x402 payment middleware and a machine-commerce operating stack. The deciding evidence in this pass was the repeated combination of zero-code buyer autopay, zero-code seller/MCP wrapping, non-custodial Solana-USDC settlement, AI-tool auto-install flows, prepaid and budget tooling, and explicit production examples.