AgentGate - Secure AI Agent Gateway with Auth0 Token Vault

A security-first gateway that manages how AI agents access third-party services on behalf of users, powered by Auth0 Token Vault.

Every token request is scoped, logged, and revocable.

The Problem

AI agents need to access services (GitHub, Slack, Google, etc.) on behalf of users. Current approaches are dangerous:

• Agents store raw API keys with full access
• No visibility into what agents do with credentials
• No way to scope or revoke access per-agent
• No audit trail of token usage

The Solution

AgentGate uses Auth0 Token Vault to provide a secure, observable layer between AI agents and user services:

Feature	Description
Scoped Access	Define exactly which APIs and scopes each agent can use
Audit Trail	Every token request logged with agent ID, action, timestamp
Step-Up Auth	Sensitive operations require additional user confirmation
Auto-Revocation	Tokens expire, rotate, and revoke automatically
Real-Time Dashboard	See what agents are doing with your credentials right now

Architecture

User ──> Auth0 Login ──> AgentGate Dashboard
                              │
                    ┌─────────┼─────────┐
                    │         │         │
                    v         v         v
              ┌─────────┐ ┌────────┐ ┌────────┐
              │ GitHub  │ │ Slack  │ │ Google │  ... Connected Services
              │ Token   │ │ Token  │ │ Token  │      via Token Vault
              └────┬────┘ └───┬────┘ └───┬────┘
                   │          │          │
                   └──────────┼──────────┘
                              │
                    ┌─────────v─────────┐
                    │   Token Vault     │
                    │   (Auth0)         │
                    │                   │
                    │  • Token storage  │
                    │  • Auto-refresh   │
                    │  • Scope mgmt     │
                    └─────────┬─────────┘
                              │
                    ┌─────────v─────────┐
                    │   AgentGate       │
                    │   Policy Engine   │
                    │                   │
                    │  • Per-agent ACLs │
                    │  • Rate limits    │
                    │  • Audit logging  │
                    └─────────┬─────────┘
                              │
              ┌───────────────┼───────────────┐
              │               │               │
              v               v               v
        ┌──────────┐   ┌──────────┐   ┌──────────┐
        │ Agent A  │   │ Agent B  │   │ Agent C  │
        │ (Code    │   │ (Slack   │   │ (Email   │
        │  Review) │   │  Bot)    │   │  Assist) │
        └──────────┘   └──────────┘   └──────────┘

How Auth0 Token Vault Powers AgentGate

1. Connected Accounts - Users authorize services once via OAuth; Token Vault manages the full token lifecycle
2. Token Exchange (RFC 8693) - Agents request scoped tokens through AgentGate, which exchanges them via Token Vault
3. CIBA Flow - Step-up authentication pushes consent requests to users for sensitive operations
4. My Account API - Users manage their connected accounts and see active sessions

Quick Start

# Clone and install
git clone <repo-url>
cd auth0-agentgate
pip install -r requirements.txt

# Configure Auth0
cp .env.example .env
# Edit .env with your Auth0 credentials

# Run
uvicorn src.app:app --reload

Configuration

Required Auth0 setup:

1. Create an Auth0 application (Regular Web Application)
2. Enable Token Vault and CIBA grant types
3. Activate My Account API with Connected Accounts scopes
4. Configure social connections (GitHub, Slack, Google)

See AUTH0_SETUP.md for detailed instructions.

API Endpoints

User Endpoints

• GET / - Dashboard with connected services and agent activity
• GET /login - Auth0 login flow
• GET /callback - OAuth callback handler
• POST /connect/{service} - Connect a new service via Token Vault
• DELETE /connect/{service} - Disconnect a service
• GET /audit - View audit trail

Agent Endpoints

• POST /api/v1/token - Request a scoped token for a service
• GET /api/v1/services - List available services for this agent
• POST /api/v1/step-up - Trigger step-up auth for sensitive ops

Testing

pytest tests/ -v

Prize Categories

• Security Model - Per-agent ACLs, step-up auth, automatic token rotation
• User Control - Real-time dashboard, granular permissions, instant revocation
• Technical Execution - Full Token Vault + CIBA + My Account API integration
• Potential Impact - Every AI agent deployment needs this security layer

License

MIT