AI Agent Runtime Security Hub
A practical hub for securing AI agents while they call tools, use memory, retrieve data, connect to MCP servers, and act through production APIs.
Prompt injection that turns into unauthorized tool execution.
MCP or plugin tools that expose files, shell commands, internal APIs, or excessive network access.
Agent memory and retrieval flows that persist or disclose sensitive context.
Built For
AI product teams moving from chat features to tool-using agents.
Security teams validating agent behavior before production rollout.
Platform and backend teams responsible for API tokens, MCP servers, and approval flows.
Use Cases
Map every tool call to identity, data access, approval, and audit requirements.
Separate prompt filtering from runtime enforcement and least-privilege tool design.
Prepare agent systems for proof-driven assessment before customer or employee rollout.
Related Content
Related Advisories
Unauthenticated Remote Code Execution via Arbitrary Command Injection in MCPHub Server Registration
MCPHub accepts attacker-controlled command and args values during server registration and spawns them through STDIO, enabling full remote code execution on the host.
Authentication Bypass via skipAuth Configuration Grants Full Admin Access in MCPHub
When skipAuth is enabled, MCPHub bypasses both authentication and admin authorization checks, allowing any unauthenticated user to access privileged API functionality.
Frequently Asked Questions
How is agent runtime security different from prompt security?
Prompt security tries to shape the model response. Runtime security validates what the agent can actually do through tools, APIs, memory, and retrieval.
When does an AI agent need runtime assessment?
When it can call tools, read sensitive data, write records, connect to MCP servers, or act with production credentials.
Need help validating this attack surface?
Talk with Eresus Security about scoped testing, threat modeling, and remediation priorities for this workflow.
Talk to Eresus