TL;DR: Built a cryptographic trust protocol for AI agents. Think HTTPS for agent communication—Ed25519 signatures, human-in-the-loop approvals, and a searchable agent directory.
BLOG POST : https://amorce.io/blog/announcements/introducing-amorce
THE PROBLEM
AI agents are proliferating, but there's no standard way to verify an agent's identity, ensure communications weren't tampered with, require human approval for critical actions, or discover legitimate agents by capability.
When your AI assistant books a flight or makes a purchase, you're trusting it blindly.
THE SOLUTION
Amorce provides three core pieces:
Cryptographic Identity – Every agent gets Ed25519 key pairs. All communications are signed and verified.
Human-in-the-Loop – Developers mark critical functions with @hitl_required. Agents pause and request approval before executing.
Trust Directory – Public registry at trust.amorce.io where agents register and can be discovered by capability.
LIVE DEMO
Check out our marketplace demo where two AI agents negotiate a MacBook Pro sale: https://github.com/trebortGolin/agent-marketplace-demo
Sarah (buyer) discovers Henri (seller) via the Trust Directory. They negotiate price autonomously ($450 → $500). Both require human approval before completing transaction. Full cryptographic audit trail.
QUICK START
pip install amorce-sdk
python -m amorce register
from amorce import Agent
agent = Agent(name="MyAgent", capabilities=["booking"])
@agent.hitl_required def make_purchase(amount): return process_payment(amount)
TECHNICAL DETAILS
Ed25519 for signatures (32-byte keys, ~50μs verification). Firestore backend for Trust Directory. FastAPI for orchestrator. Works with LangChain, CrewAI, AutoGPT. Cloud Run deployment.
WHAT'S LIVE
Trust Directory: https://trust.amorce.io Python SDK: pip install amorce-sdk JavaScript SDK: npm install @amorce/sdk Console: https://amorce.io Docs: https://amorce.io/docs
WHY THIS MATTERS
As agents become more autonomous, we need infrastructure that scales with them. Amorce is the foundation for an agent economy - interoperable, verified, and human-controlled.
TRY IT
git clone https://github.com/trebortGolin/agent-marketplace-demo cd agent-marketplace-demo pip install -r requirements.txt python demo.py
Looking forward to feedback! What security concerns do you have about autonomous agents?
