As AI agents evolve from simple chat interfaces into fully distributed workloads, the security landscape around them must evolve as well. Multi-agent workflows, multi-workspace architectures, MCP servers, and tool-driven automation introduce new classes of risks that traditional firewalls were never designed to detect.

Today, we’re pleased to announce our partnership with AgentX, a leading platform for building and deploying multi-agent systems across enterprises.

Together, we’re bringing governance and visibility into environments where AI agents act autonomously, execute complex toolchains, and access critical enterprise data.

Why This Partnership Matters

AgentX deployments now rely on agents pulling data from:

• Private Databases (MCP Servers)

• SaaS Tools (CRMs, Ticketing)

• Unstructured Documents (RAG)

These behaviors are powerful - but they open the door to data leakage.

Golf introduces a new type of firewall that validates the traffic returning to the agent:

• Sanitizing Tool Outputs: Ensuring that credit cards, API keys, and PII are redacted from responses in real-time.

• Blocking Malicious Payloads: detecting "jailbreak" attempts hidden inside external data sources.

• Auditing Data Consumption: proving exactly which documents and records were fed into the AI model.

Together with AgentX, we’re establishing the first full-stack approach to securing modern multi-agent systems.

Looking Ahead

This partnership accelerates our mission: Protecting AI systems at the protocol, not just network, level.

AgentX will standardize on Golf’s firewall for enterprise deployments, giving organizations a safer foundation for building advanced agent architectures powered by MCP, RAG, and distributed tools.

The next era of AI requires a new cybersecurity layer - and that’s what we are building.