Aegis is an AI agent governance platform that sits between your coding agents and GitHub. When an agent, such as Claude Code, Cursor, or VS Code Copilot, calls a tool that touches your repositories, Aegis intercepts the request over the Model Context Protocol (MCP) and evaluates it against your configured policies before anything executes. You decide which actions go through automatically, which get rewritten to a safer path, which are blocked outright, and which pause for a human to review. No code changes are required in your agent or your repositories.Documentation Index
Fetch the complete documentation index at: https://docs.runaegis.co/llms.txt
Use this file to discover all available pages before exploring further.
Why governance matters
AI coding agents can push commits, open pull requests, modify workflows, and merge branches, often faster than a human reviewer can keep up. Without a control layer, a single misconfigured prompt or an overly broad instruction can cause a force-push to a protected branch, merge a PR without review, expose secrets in a commit, or trigger a broken deployment pipeline. Aegis ensures every consequential move stays intentional, auditable, and approved.Key capabilities
Policy engine
Evaluate every agent tool call against a set of rules. Policies can Allow, Deny, Rewrite, or require human Approval, and take effect the moment you save them.
Human-in-the-loop approvals
Route sensitive actions to the Aegis dashboard for one-click approval or denial. The agent pauses and waits for your decision before continuing.
Immutable audit trail
Every tool call, decision, and outcome is recorded in a tamper-proof log. You can filter by agent, repository, tool, or time range.
Rooms
Rooms scope governance to a specific repository. Each room has members with roles (OWNER, ADMIN, DEVELOPER) and per-tool access controls configured per role.
Freeze windows
Block all agent activity across one or more repositories during deployments, incidents, or planned maintenance, with a single toggle.
Real-time dashboard
The dashboard auto-refreshes every 30 seconds so you always see the current state of agent activity, pending approvals, and decision counts.
Who Aegis is for
Aegis is built for engineering teams that use AI coding agents in their day-to-day workflow. If your team is already running Claude Code, Cursor, or VS Code Copilot against production repositories, Aegis gives you the visibility and control you need to do that safely, without slowing agents down for routine tasks or losing oversight on the moves that matter.How it connects
Aegis exposes an MCP SSE endpoint athttps://app.runaegis.co/sse. You add a single JSON snippet to your agent’s MCP configuration pointing at that URL, and Aegis begins intercepting tool calls immediately. There is nothing to install in your repositories and no agent code to modify.
Aegis uses the Model Context Protocol over Server-Sent Events (SSE). Any MCP-compatible agent can connect, not just Claude Code, Cursor, and VS Code Copilot.
Next steps
Quickstart
Create your account and connect your first agent in under five minutes.
How Aegis works
Understand the full request lifecycle from agent tool call to GitHub action.