Aegis — AI Agent Governance
Govern which tools your AI agent may call using YAML policies, human approval gates, and audit logs before risky actions hit production.
Overview
io.github.Acacian/aegis is a MCP server for the Ship phase that enforces YAML policy, approvals, and audit logging on AI agent tool calls.
What is this MCP server?
- YAML-defined policies for AI agent tool invocations
- Approval gates before sensitive or disallowed tool calls proceed
- Audit logging for accountability across agent sessions
- PyPI package agent-aegis with optional AEGIS_POLICY_PATH env
- stdio MCP transport for local agent hosts
- Registry version 0.1.7 (PyPI package 0.1.6 in metadata)
- Policy configuration via AEGIS_POLICY_PATH environment variable
- stdio MCP transport via PyPI identifier agent-aegis
Community signal: 9 GitHub stars.
What problem does it solve?
Powerful coding agents can execute destructive or costly tools faster than you can notice, with no single place defining what is allowed.
Who is it for?
Indie builders running Claude Code or similar agents against real repos who want explicit tool governance before sharing workflows or shipping automation broadly.
Skip if: Static application penetration testing or teams with no agent tool access who only need traditional SAST dashboards.
What do I get? / Deliverables
After registration, tool calls flow through Aegis so policy violations are blocked or held for approval, with an audit trail you can review.
- Enforced tool-call policy with optional human approval steps
- Persistent audit logging of agent tool attempts and decisions
- Central governance layer reusable across agent projects
Recommended MCP Servers
Journey fit
Tool-call governance is a Ship concern when agents can touch files, networks, or payments and you need guardrails before wide release. Security is the canonical shelf because Aegis enforces policy, approvals, and auditing on agent actions rather than building product features.
How it compares
Policy and approval MCP for agent tools, not a network pentest suite or a secrets scanner alone.
Common Questions / FAQ
Who is io.github.Acacian/aegis for?
Solo builders and small teams operating AI coding agents who need YAML policies, approval gates, and audit logs on tool usage.
When should I use io.github.Acacian/aegis?
Use it during Ship security hardening and ongoing agent operation whenever tools can modify infrastructure, data, or external services.
How do I add io.github.Acacian/aegis to my agent?
Install agent-aegis from PyPI, set AEGIS_POLICY_PATH to your YAML policy file, add the stdio MCP server in your agent config, and route tool calls through Aegis per project docs.