mcpdirectory.dev
9,129 servers · updated 23 Sept

Aegis MCP server

community Active last commit 25 d 
{
  "mcpServers": {
    "acacian-aegis": {
      "command": "uvx",
      "args": [
        "agent-aegis"
      ]
    }
  }
}
file .mcp.json (in your project root)install uvx agent-aegisversion 0.1.6

What it does

Aegis is a governance layer for AI agents that provides policy-based control over tool calls across 12 agent frameworks through a single unified API. It enforces governance primitives like prompt-injection blocking, PII masking, policy enforcement, trust delegation, and audit logging without requiring code changes to agent implementations.

  • Define YAML policies once and apply them uniformly across LangChain, CrewAI, OpenAI Agents, and other supported frameworks without modifying agent code.
  • Block tool calls that violate security policies, mask sensitive data, and enforce rate limits or cost budgets across all agent invocations.
  • Generate tamper-evident audit logs of agent actions, including delegation chains and justification gaps, for compliance and security review.

Tools this server exposes

ToolDoes
auto_instrumentAutomatically instruments installed agent frameworks with governance hooks in a single call.
PolicyDeclarative YAML-based rules engine that matches tool calls and applies risk-based approvals.
ActionUnified representation of tool, LLM, HTTP, or MCP calls across all supported frameworks.
ActionClaimTripartite structure capturing declared intent, Aegis-assessed properties, and delegation chains.
GuardrailsDeterministic checks for prompt injection, PII, and other security patterns.
AuditEventTamper-evident audit trail of agent actions and governance decisions.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.