What is AgentPort?
AgentPort is an open-source security gateway designed to protect AI agents and autonomous systems. It acts as a control layer between agents and external services, monitoring and filtering requests to prevent unauthorised actions, data leaks, or malicious behaviour. The tool is intended for developers and organisations deploying AI agents in production environments who need to enforce security policies without sacrificing functionality. By running locally or self-hosted, AgentPort gives teams direct control over agent permissions and audit trails, making it suitable for regulated industries or high-security applications where third-party agent management is not acceptable.
Key Features
Request filtering
Inspect and control what external calls agents can make
Policy enforcement
Define rules for agent behaviour based on custom security policies
Audit logging
Track all agent actions and external requests for compliance and debugging
Open-source codebase
Full transparency and ability to customise the gateway for specific needs
Self-hosted deployment
Run on your own infrastructure with no external dependencies
Pros & Cons
Advantages
- Complete transparency as an open-source project; you can review and modify the code
- No vendor lock-in; deploy on your own servers or infrastructure
- Granular control over agent permissions and behaviour without relying on third-party services
- Useful for regulated industries where data must remain within your control
Limitations
- Requires technical expertise to deploy and maintain in production
- Limited commercial support compared to proprietary solutions; community-driven development may move slower
Use Cases
Protecting autonomous agents in enterprise environments from making unintended external API calls
Enforcing compliance policies for AI systems handling regulated data
Auditing and monitoring agent behaviour in financial or healthcare applications
Preventing data exfiltration by restricting where agents can send information
Testing agent security policies in development before production deployment