Description
Warrant is a security tool designed to manage the capabilities of AI agents through cryptographically signed policies. These policies are deny-by-default, tamper-proof, and can be set up in just ninety seconds. Warrant addresses the problem of AI agents having unrestricted access, which can lead to issues such as data deletion, secret exfiltration, and malware installation. By using Warrant, users can define exactly what their AI agents can do, ensuring that unauthorized actions are blocked before execution.
Read More
Warrant works with a wide range of AI frameworks, including Claude Code, Codex, Copilot, and LangChain. It operates by installing manifests for various tools like git, npm, and pip, and locks the policy with a shell guard hook. This setup allows agents to run normally while blocking denied commands. Warrant can also lock down entire machines, providing security for agents running on dedicated hardware.
The tool is built for real security, featuring an open manifest standard, Ed25519 signing, monotonic versioning, and a tamper-evident audit trail. It also offers elevation and override capabilities for temporary unrestricted access. Warrant's package security feature blocks known-malicious packages, enhancing overall security.
Warrant is open source and offers a free tier with core security tooling. An enterprise version adds operational features like central policy administration, centralized audit logs, and role-based access control. The tool is suitable for teams running agents at scale, providing a comprehensive security solution for AI agent management.
- CategoryAI Cybersecurity Tools
- Pricing ModelFreemium
- DeploymentSelf-hosted
- PlatformsAny AI framework
Warrant's Core Features
Cryptographically signed policies
Deny-by-default access control
Tamper-proof security
Ninety-second setup
Works with any AI framework
Open manifest standard
Ed25519 signing
Monotonic versioning
How to use Warrant?
Configure: Define capability policies
Use: Install manifests and lock policies
Enforce: Run agents with security in place
Optimise: Adjust policies as needed
Warrant's Use Cases
- AI Agent Security
- Access Control
- Policy Management
- Machine Lockdown
- Package Security
