Downloads · 30 days
0
A2SPA_The_Missing_Security_Layer_for_AI_Agents
devincapriola/A2SPA_The_Missing_Security_Layer_for_AI_Agents
A2SPA_The_Missing_Security_Layer_for_AI_Agents is a machine learning model from devincapriola. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Agent-to-Secure Payload Authorization | Get Started | Documentation
Access
Public
Updated Feb 23, 2026
Repo size
—
Likes
1
Public
Repo makeup
Click a slice to open those files.
.md4.2 KB · 74%
At a glance
- Access
- Public
- Created
- Feb 23, 2026
- Updated
- Feb 23, 2026
- SHA
- 359eb8ac
Model card
From the Hugging Face model README
Agent-to-Secure Payload Authorization | Get Started | Documentation
The Problem
Every major AI agent framework today — LangChain, AutoGen, CrewAI, MCP, AWS Bedrock — treats incoming payloads as legitimate by default.
Orchestration routes them. Tool schemas validate their shape. Sandboxing contains execution. Guardrails check the output. But nobody cryptographically verifies that a payload was:
- Actually sent by the agent who claims to have sent it
- Unmodified in transit
- Not replayed from a previous session
- Authorized to be sent by that agent
This is the Payload Trust Gap — and it's what enables every serious agent exploit including prompt injection, agent hijacking, and replay attacks.
What A2SPA Does
A2SPA is a cryptographic control layer that sits at the execution boundary of your agent stack (Layer 5 — between orchestration and execution) and enforces verification on every single payload before it runs.
Orchestration Layer
↓
Tool Schemas
↓
Sandboxing
↓
Permissions
↓
Guardrails
↓
Logging
↓
━━━━━━━━━━━━━━━━━━━━━━━━━
PAYLOAD TRUST GAP ⚠️
(all layers above assume
the payload is legitimate)
━━━━━━━━━━━━━━━━━━━━━━━━━
↓
┌─────────────────────┐
│ A2SPA CONTROL │
│ LAYER │
│ Sign → Verify → │
│ Log │
└─────────────────────┘
↓
EXECUTION BOUNDARY
(Arms, Drones, Wallets,
Tools, Robots)
Features
| Feature | Description |
|---|---|
| 🔐 Cryptographic Signatures | Every payload SHA-256 hashed and signed with the agent's private key |
| 🔄 Nonce Replay Protection | Unique nonce per request with 24-hour TTL storage |
| 🔒 Agent Permission Mapping | Granular per-agent send/receive control with instant ON/OFF toggle |
| ✅ Identity Binding | Every action tied to a verified agent identity, not just an API key |
| 📋 Tamper-proof Audit Trail | Complete interaction logging with CSV export |
| 📊 ROI Tracking | Time saved and dollar value generated per agent |
How It Compares
| Security Feature | A2SPA | MCP | A2A | ACP | ANP | LangChain | AWS Bedrock |
|---|---|---|---|---|---|---|---|
| Payload Signing | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Nonce/Replay Protection | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Permission Mapping | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ |
| Audit Logging | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
| Zero-Trust Ready | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
What A2SPA Enforces on Every Payload
Authenticity — The payload was signed with the sending agent's private key. SHA-256 hash + signature verification on every request.
Authorization — The sending agent has permission to send this type of payload. Granular per-agent permission mapping with instant on/off toggle.
Non-repudiation — Tamper-proof record that this specific agent sent this specific payload at this specific time.
Integrity — Any modification in transit invalidates the signature.
Replay Protection — Unique nonce with 24-hour TTL. A captured payload cannot be replayed.
Identity Binding — Every action tied to a verified agent identity, not just a session token.
Pricing
$0.01 per verification — pay-as-you-go, no minimums, no upfront commitment.
Framework Compatibility
A2SPA is framework-agnostic. It works with any agent framework including LangChain, AutoGen, CrewAI, and custom implementations.
Get Started
About
A2SPA was invented by Jonathan, Devin & Gavin Capriola. Patent pending.
The cryptographic firewall for AI agents.
- Created
- Feb 23, 2026
- Updated
- Feb 23, 2026