Home / Technology / AI Agents Get Guardrails: Control Your Digital Tasks
AI Agents Get Guardrails: Control Your Digital Tasks
23 Oct
Summary
- New tools ensure AI agents stay within set operational limits.
- Products control AI actions in software development and finance.
- Decisions are recorded, creating a verifiable history of AI actions.
A digital asset infrastructure company, AEREDIUM, has introduced new products to manage the increasing autonomy of AI agents. The Chair and MCP Connector, available on the aeredium.ai platform, are designed to place controls between AI agents and the systems they interact with.
These tools employ a policy engine to evaluate and approve or deny proposed AI actions. Organizations can establish rules to permit, block, or flag actions for human review. Every decision is digitally signed and logged, creating a verifiable record. Albert Dadon, CEO of AEREDIUM, stated that the goal is to enable AI deployment without losing control over its operations.
The Chair specifically targets software development, monitoring AI coding agents' interactions with development environments. It checks actions like file modifications and code commits against company policies. The tool supports various coding agents and allows policy configuration through a questionnaire, offering a 30-day free trial.
MCP Connector applies similar controls to financial activities, combining a digital asset wallet with a policy engine. This allows AI agents to execute payments or trades within defined permissions. Each agent has its own wallet and a designated human owner who approves mandates. The system supports stablecoin transfers and trading on designated exchanges, with transaction records retained for seven years.
