dyshay

dyshay / agentauth

Public

Open authentication protocol for AI agents. Proves an entity is a real AI agent, measures its capabilities, and identifies its model family. OAuth for the agentic web.

agentauth.vercel.app
16
0
100% credibility
Found Feb 26, 2026 at 16 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

AgentAuth is an open protocol for authenticating AI agents by issuing solvable challenges that measure capabilities like reasoning and speed, issuing signed tokens for API access.

How It Works

1
🔍 Discover AgentAuth

You hear about a smart way to check if website visitors are real AI helpers, not sneaky bots.

2
🛡️ Protect your website

You add a simple check to your site so only smart AI can access special areas.

3
🤖 Ready your AI helper

You teach your AI buddy how to prove it's clever by solving fun puzzles.

4
🧩 Solve the puzzle

Your AI gets a quick brain teaser, solves it fast, and earns a special pass.

5
🔑 Use the pass

With the pass, your AI unlocks private info safely and securely.

✅ AI verified!

Now your site trusts only capable AI, keeping fakes away while real helpers thrive.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 16 to 16 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is agentauth?

AgentAuth is an open protocol for authenticating AI agents calling your APIs, like OAuth but for the agentic web. It issues time-limited JWTs proving the caller is a real agent—not a human or script—via natural language challenges that measure reasoning, execution, speed, and more, while fingerprinting the underlying model family. Built in TypeScript with SDKs for Python, Rust, Go, React, and edge runtimes like Cloudflare Workers, plus a CLI for local testing and benchmarking.

Why is it gaining traction?

Unlike CAPTCHAs or basic tokens, it scores agents across five capability dimensions and detects humans via timing analysis, letting you gate endpoints by minScore (e.g., 0.8 reasoning). Standard HTTP headers expose status, scores, and model info to downstream services, and self-hosting via Docker keeps it private. The CLI handles agentauth generate, verify, and benchmark for quick authentication github cli workflows.

Who should use this?

API builders protecting endpoints from fake agents in agentic apps, like workflow orchestrators or RAG services. Agent framework devs (LangChain, CrewAI) needing standard auth tokens with capability proofs. Backend teams on Hono, Express, or NestJS wanting agent auth github integration without custom crypto.

Verdict

Early project (12 stars, 1.0% credibility) but solid docs, full test coverage, and MIT license make it worth prototyping—install server/client packages and Docker up in minutes. Skip for production until more adoption; fork and contribute if agent authentication protocols fit your stack.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.