Pre-execution authority enforcement for AI agents

---

Overview

Caracal is a pre-execution authority enforcement system for AI agents and automated software operating in production environments. It exists at the boundary where autonomous decisions turn into irreversible actions—such as API calls, database writes, or system triggers.

By enforcing the principle of explicit authority, Caracal ensures no action executes without a cryptographically verified, time-bound mandate issued under a governing policy.

---

Community

GitHub's Open Source Friday

More coming soon

---

Installation & Setup

Quickstart

caracal up # Setup

caracal cli # Containerized CLI
# OR
caracal flow # Terminal UI

Command Reference

caracal up         # Pull images, create network/volumes, start postgres+redis+mcp
caracal down       # Stop stack and remove services
caracal cli        # Run full Caracal CLI inside container
caracal flow       # Run TUI inside container
caracal logs -f    # Tail runtime logs
caracal reset      # Down + remove volumes (full local reset)
caracal purge      # Completely remove Caracal containers, data, networks, images, and local state

Migration and Cleanup

caracal migrate repo-to-package
caracal workspace delete <workspace-name> --force
caracal reset
caracal purge --force

---

Contributing

See CONTRIBUTING.md for setup, workflow, tests, and pull request standards.

---

Citation

Caracal is an open-source framework for pre-execution authority enforcement for AI agents controlling delegated actions, with real-time revocation and immutable proof.

If this project contributes to your research, product, or derivative systems, please consider citing it to help us advance trustworthy AI security research.

@software{madhuwala2026caracal,
  author    = {Madhuwala, Ryan and Garudex Labs},
  title     = {Caracal: Authority Enforcement Framework for AI Agents},
  year      = {2026},
  publisher = {GitHub},
  url       = {https://github.com/garudex-labs/caracal}
}

---

License

Caracal is open-source software licensed under the Apache-2.0 License. See the LICENSE file for details.

Developed by Garudex Labs.