Summary

Agent Control addresses the challenge of enforcing runtime guardrails for AI agents at scale. It provides a centralized control plane to govern agent behavior, preventing prompt injections, PII leakage, and other risks without modifying agent code. This solution benefits teams by enabling centralized safety configurations, runtime management via API/UI, and rapid updates to protections without redeployments.

How It Works

The system operates as a centralized control layer that evaluates agent inputs and outputs against configurable rules. Core components include controls (defining rules), selectors (specifying data paths like output), and evaluators (applying logic like regex, SQL, or custom functions). This approach decouples safety logic from agent code, allowing for central management and updates through an API or UI, enhancing security and maintainability.

Quick Start & Requirements

• Prerequisites: Docker, Python 3.12+.
• Installation:
  • Server: curl -L https://raw.githubusercontent.com/agentcontrol/agent-control/refs/heads/main/docker-compose.yml | docker compose -f - up -d
  • SDK: uv pip install agent-control-sdk (Python).
• Setup Flow: Start server -> Install SDK -> Wrap agent functions with @control() -> Register agent -> Create controls via UI or SDK.
• Links: Agent Control Website, Docs, [Quickstart](https