Effortlessly add authorization and permissions to your AI agent!

With just a few lines of code. Built for fast moving AI developers!

Get Started

See Examples

Without DeepSecure

No AI agent identity

No ephemeral credentials

No secure access to tools

No access control

No delegation and tracking

Production blockers

With DeepSecure

AI Agents get cryptographic identity

Ephemeral and runtime credential injection

Easily add secure tool calling

AI Agents with fine grained authorization

User <> AI Agent <> AI Agent delegation and audit

Production and enterprise ready

HOW IT WORKS

Start Building in Minutes

Go from install to secure agent in three steps. No complex config, just clean integration.

Step 1: Install & Configure

Install the deepsecure library
Start the services
Configure for CLI and SDK

# Install the library from PyPI
pip install deepsecure

# Clone the repository and start the services
git clone https://github.com/DeepTrail/deepsecure.git
cd deepsecure
docker-compose up -d

# Configure the URL for the Control Plane (for admin tasks via CLI)
deepsecure configure set-control-plane-url http://localhost:8000

# Configure the URL for the Gateway (for agent runtime via SDK)
deepsecure configure set-gateway-url http://localhost:8001

Step 2: Give Your AI Agent a Secure Identity

Gives your agent a unique, cryptographic identity. DeepSecure automatically handles fetching short-lived, secure credentials from now on.

# Before: Unsafe, hard-coded keys
# import os
# from openai import OpenAI
# client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))

# After: Secure, auditable agent identity
import deepsecure

client = deepsecure.Client()

# This single line creates a unique identity for your agent
# and prepares it for secure, audited access.
agent = client.agent("my-financial-analyst", auto_create=True)

# Now your agent can access secrets without ever touching the raw keys
secret = client.get_secret(agent_id=agent.id, secret_name="openai-api")

Step 3: Delegate Tasks Securely Between Agents

Enable agents to securely delegate tasks to each other with temporary, fine-grained permissions. This creates a secure, auditable chain of command for complex workflows.

# In your "Manager" agent's code:
# Create a temporary, restricted token for another agent
delegation_token = client.delegate_access(
    delegator_agent_id="manager-agent-id", 
    target_agent_id="finance-agent-id", 
    resource="financial-data", 
    permissions=["read"], 
    ttl_seconds=1800  # Expires in 30 minutes
)

# In your "Finance" agent's code:
# Use the delegated token to perform the specific task
client.access_resource_with_delegation(
    token=delegation_token,
    resource="financial-data"
)

# The finance agent can only access what was delegated, for a limited time.
# All actions are logged and auditable.