deepsecure

DeepSecure provides a zero-trust security control plane for AI agents. It assigns cryptographic identities to each agent (Ed25519), enforces fine-grained access policies, supports secure delegation between agents, and maintains an audit trail of actions. The system comprises a Control Plane that manages agents, policies, and credentials, and a Gateway (Data Plane) that enforces policies and injects ephemeral secrets as needed. You can interact with it via the DeepSecure Python package and its CLI to create agents, configure the control plane URL, and retrieve secrets for use in AI workflows. Typical usage involves creating an agent identity, obtaining ephemeral credentials, and then using those credentials to call third-party APIs or internal services under policy controls and with full visibility into access events.

Prerequisites:

• Python 3.9+
• pip (Python package installer)
• Docker and Docker Compose (for running backend services locally)

Installation steps:

1. Install the DeepSecure Python package

pip install deepsecure

2. Run backend services (recommended via Docker Compose)

# Clone and set up the project (example commands)
# git clone https://github.com/DeepTrail/deepsecure.git
# cd deepsecure

# Start both Control Plane and Gateway with Docker Compose
docker-compose up -d

# Verify services are running
docker-compose ps

This will start the Control Plane at http://localhost:8000 and the Gateway at http://localhost:8001, along with a PostgreSQL database for persistence.

3. Configure the DeepSecure CLI to point to your control plane

# Set the control plane URL
deepsecure configure set-url http://localhost:8000

# Verify connection
deepsecure health

4. Quick validation

# Check version
deepsecure --version

# Create a test agent
deepsecure agent create --name "test-agent"

Tips and notes:

• Ensure Docker and Docker Compose are installed if you plan to run the backend locally via Docker for a quickstart.
• The control plane URL is central; point your CLI to the correct instance (local vs. production) using deepsecure configure set-url.
• Secrets are injected by the Gateway and policy enforcement happens automatically when agents attempt to access restricted APIs.
• Regularly review the audit logs for access patterns and potential policy improvements.
• If you encounter connectivity issues, verify that the backend services are healthy and the database is accessible.
• No hardcoded credentials should be stored in the agent code; rely on ephemeral credentials managed by the control plane.