DEV Community

Wai Yip, WONG
Wai Yip, WONG

Posted on • Originally published at github.com

Launching ANCP — Secure Login for AI Agents Without VPN or OAuth

🚀 Launching ANCP — Secure Login for AI Agents Without VPN or OAuth

AI agents are evolving from tools into autonomous reasoning actors — capable of reading, writing, and acting on sensitive data. But today’s authentication systems were built for humans with browsers, not intelligent agents with goals.

That’s why I’m launching ANCP — the Agent-Native Challenge Protocol.

It’s a new kind of login protocol built specifically for LLMs like ChatGPT, Gemini, Claude, and future autonomous agents.


🔐 What Makes ANCP Different?

ANCP lets an AI agent authenticate and access backend infrastructure without:

  • ❌ VPN tunnels
  • ❌ OAuth 2.0 or browser redirects
  • ❌ Static API keys
  • ❌ Service accounts or secret-sharing

Instead, it uses:

  • PGP-based challenge–response login
  • Zero-trust, stateless session negotiation
  • AI-readable .well-known/ancp.discovery.json endpoint metadata
  • Prompt-native reasoning over identity

🧠 Why It Matters

Today’s LLMs are capable of:

  • Making API calls
  • Generating SQL queries
  • Accessing CRMs or HR systems
  • Recommending product launches, policy decisions, or financial strategies

But if the only login method available is:

  • “Paste this OAuth token in a browser...”
  • …the agent becomes blocked or dangerously over-permissioned.

ANCP makes identity a reasoning-aligned interaction.

The agent proves it understands what it's doing — cryptographically and semantically.


📦 GitHub Repository

👉 https://github.com/waiyip000/agent-native-auth

Includes:

  • Full whitepaper (whitepaper.docx)
  • SHA256 validation hash
  • .well-known/ancp.discovery.json sample
  • PGP digest tool (digest.py)
  • Attribution-secured LICENSE

📄 Whitepaper SHA256 Digest

To verify authorship and originality of the protocol:

SHA256 (whitepaper.docx):
407f2e29bc6ba9bb8710cd43da58401d056bd1d1cb7c0dadeb6729a99a9c4c36
Enter fullscreen mode Exit fullscreen mode

🧑‍💻 Created and authored by

Wai Yip, WONG

🔗 LinkedIn

💻 GitHub


📣 Call to Action

I’m inviting:

  • OSS contributors
  • LLM security architects
  • AI infra developers
  • Zero-trust security teams

…to collaborate, test, and build on ANCP.

Let's make secure, intelligent infrastructure agent-native.

ANCP #LLM #AI #ZeroTrust #Security #Authentication #PGP #PromptNative #OpenSource

Top comments (0)