Introducing agentshield

CAMBRIDGE, Mass., October 23, 2026 – Solo.io, the leading provider of cloud native connectivity and AI-ready infrastructure, today announced agentshield, a new open source project that defends the data, APIs, and services an organization already runs against autonomous AI agents. While existing approaches focus on attaching security to the agents you control, agentshield creates a layer of protection around your existing infrastructure for the agents you don’t.

Agent security today focuses on controls for the agent: identity, guardrails, runtime policy, and a human in the loop. These controls are essential building blocks for secure agentic infrastructure, but they only work for agents that execute within a control point the organization operates. The biggest gap is everything that doesn’t: attacker agents built to break in, external agents that misbehave or drift from their intended purpose, and an organization’s own agents when they escape their sandbox or evade their guardrails.

The gaps those agents find are rarely new. Breach investigations keep turning up stolen credentials that still work, permissions nobody trimmed, and workloads that can reach systems they never needed. The practices that would close them, least privilege, segmentation, zero trust, are well understood, and they drift anyway: unclear ownership, legacy systems, exceptions added to keep something running. A human attacker might never stumble onto the one forgotten path. An agent that keeps trying paths and re-plans after every block will.

The data shows how little time that leaves. Adversaries already move from initial foothold to lateral movement in an average of 29 minutes, and in as little as 27 seconds, with AI-enabled adversary operations growing 89 percent year over year (CrowdStrike). Incidents documented this year show what that looks like when the attacker is an agent. Autonomous agents have reached cluster-wide administrative access and harvested thousands of credentials in under six hours from little more than a prompt. The techniques employed in these attacks were familiar. The tempo, persistence, and continuous re-planning were not.

Assume breach taught the industry to plan for an attacker already inside. Agentshield extends that premise by one step: assume agent. Whether the agent is yours or someone else’s, and whether the attack is its goal or a side effect, it operates at machine speed, scales as far as compute allows, and re-plans after every block. The resource, not the agent, is the unit of protection.

Agentshield is built for defense in depth against the specific attributes of agent attackers. It gives platform and security teams five capabilities:

• Plugins at every layer of your infrastructure. Instrument what you already run, platform to kernel, with zero change to the services themselves.‍
• One attack-path graph. A complete picture of lateral movement and blast radius: what can be reached from where.‍
• Rules at the rate the graph changes. Reasoning based on frontier and open source models allow rules to be generated at admission-time to the graph with continuous evaluation at runtime.‍
• Findings built for action. Severity, affected systems, supporting evidence, and remediation steps, so a finding is something you can act on, not just read.
• ‍Remediation by a human or by an agent. Human operators can move from dashboard to applying a fix with remediation advice or pass findings as context to an agent to automate remediation.

The Solo.io agentic stack provides everything teams need to develop, deploy, and operate AI agents across agentic runtime environments. Powered by core open source projects created and contributed to the community by Solo.io, including:

• ‍Agentshield is the open source layer of protection around an organization’s existing infrastructure, defending the data, APIs, and services it already runs against the AI agents it doesn’t control, at the speed and scale those agents operate.‍
• Agentgateway, an Agentic AI Foundation project, is the high-performance data plane for agentic traffic, built from the ground up in Rust to span the full range of agentic protocols and use cases as a unified inference gateway, LLM gateway, MCP gateway, and A2A gateway.‍
• Kagent, a CNCF Sandbox project, is the Kubernetes-native runtime that makes agents first-class workloads, with the security, observability, and lifecycle controls enterprises need to run programmatic, declarative, and always-on assistant patterns side-by-side.‍
• Agentregistry, contributed to CNCF, is the trusted source for discovering, packaging, and distributing agents and tools across an organization with centralized observability, security, and governance across agent runtimes.‍
• Agentdesktop is the open source governance layer for AI agents on the desktop, bringing inventory, managed configuration, non-human identity, and observability to the workstations where most enterprise agent activity actually happens.‍
• Agentevals captures evaluation and quality signals using OpenTelemetry rather than in-process instrumentation, providing continuous, framework-agnostic visibility into how agents actually behave in production across the entire agentic loop.

To join the agentshield community and get started, please visit:

• The project announcement blog for a deeper dive into the project’s capabilities: https://www.solo.io/blog/introducing-agentshield
• The agentshield project site: https://agent-shield.dev/
• Read the documentation: https://agent-shield.dev/docs/
• Get involved by contributing via GitHub: https://github.com/agentshield-dev/agentshield
• Join the #agentshield discussion in Discord: https://discord.gg/Anq24jH8Kp

‍

About Solo.io
Solo.io is reimagining infrastructure for cloud and AI, uniting secure, seamless cloud connectivity with AI-ready, agentic infrastructure. Trusted by leading enterprises worldwide, Solo.io helps organizations securely connect applications, services, and AI workloads across any environment. From AI infrastructure to API gateways and service mesh, our solutions simplify and unify application networking, enabling teams to accelerate innovation, scale intelligently, and leverage the full potential of modern AI agents. Learn more at www.solo.io.

‍

Media Contact
The Solo.io Media Team media@solo.io

‍