vaultak-sentry 0.2.0

Zero-code runtime security daemon for AI agents

Vaultak Sentry

Zero-code runtime security for AI agents.

Vaultak Sentry monitors any AI agent at the OS level without requiring code changes. Install once, run any agent through Sentry, and get full behavioral monitoring, policy enforcement, and automatic violation response visible in your Vaultak dashboard.

Part of the Vaultak platform:

• Vaultak Core (pip install vaultak): SDK for developers who want deep code-level integration
• Vaultak Sentry (pip install vaultak-sentry): Daemon for teams governing existing agents without code changes

Installation

pip install vaultak-sentry

Quick Start

# Authenticate
vaultak-sentry auth --api-key vtk_your_key_here

# List available policy templates
vaultak-sentry policy list

# Create a policy for your agent
vaultak-sentry policy create --agent-id my-agent --template data-pipeline

# Run any agent with zero code changes
vaultak-sentry run --name my-agent python my_langchain_agent.py
vaultak-sentry run --name my-agent node my_agent.js

# Attach to an already running process
vaultak-sentry attach 12345 --name my-agent

# Check status
vaultak-sentry status

Policy Templates

Template	Description
data-pipeline	Read-only data processing agent
coding-agent	Software engineering agent with code execution
customer-support	Customer support agent with CRM access
research-agent	Web research agent with broad read access
hipaa-agent	HIPAA-compliant healthcare agent
strict	Maximum restriction policy

Response Modes

Mode	Behavior
alert	Log violations and notify dashboard. Agent keeps running.
pause	Stop the agent immediately on violation.
rollback	Attempt to reverse recent actions then stop the agent.

What Sentry Monitors

• File system access (reads, writes, deletes)
• Network connections (outbound and inbound)
• Child process spawning
• CPU and memory spikes
• Database connections via port detection

Advanced Usage

# Strict mode with custom blocked resources
vaultak-sentry run \
  --mode rollback \
  --block "prod.*" "*.env" "*.key" \
  --max-risk 0.7 \
  --name my-production-agent \
  python agent.py

# Disable specific monitors
vaultak-sentry run --no-net --no-proc python agent.py

# Custom policy inline
vaultak-sentry run \
  --allow "/data/readonly/**" "/tmp/**" \
  --block "*.env" "prod.*" \
  --allow-hosts "api.openai.com" \
  python agent.py

Links

• Website: https://vaultak.com
• Dashboard: https://app.vaultak.com
• Docs: https://docs.vaultak.com