# getaegis/aegis [Health: Active]

**Category:** 🔒 Security  
**Repository:** https://github.com/getaegis/aegis  
**GitHub Stars:** 14  
**npm Downloads (last month):** 440  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/getaegis-aegis

## Description
Credential isolation proxy for AI agents. Injects secrets at the network boundary with domain restrictions, agent authentication, and audit logging. No SDK required — works as a transparent HTTP proxy or MCP server.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "aegis": {
    "command": "npx",
    "args": ["-y","@getaegis/cli"]
  }
}
```

## Documentation

## What getaegis/aegis does

getaegis/aegis places a controlled boundary between an AI agent and the APIs the agent needs to call. Credentials are stored in an encrypted vault and added to outbound requests by the proxy rather than being passed to the agent. The repository describes support for AES-256-GCM encrypted credential storage, OS keychain storage by default, and file fallback across macOS, Windows, and Linux.

The proxy can restrict each credential to an allowlist of domains. It also supports agent identities, credential grants, rate limits, request body inspection, and audit records for both permitted and blocked requests. Optional policy files can apply method, path, rate, and time-of-day rules. A web dashboard, Prometheus metrics endpoint, webhook alerts, RBAC, and separate vaults are also documented features.

## How it works

In HTTP proxy mode, an agent sends a request to the local Gate process using a service path. A target-host header can identify the upstream domain when needed. Gate authenticates the agent when authentication is enabled, checks the destination against the credential’s domain restrictions, inspects the request, injects the selected secret, forwards the request, and records the result.

The MCP mode exposes the same security pipeline without requiring the agent to make HTTP calls directly. The MCP configuration points to the CLI’s `mcp serve` command, while the vault remains outside the agent’s configuration. This makes the getaegis/aegis MCP server suitable for MCP-compatible hosts that need credentialed API access without embedding secrets in environment variables.

## Setup and configuration

Install the CLI with npm, Homebrew, or Docker. The documented runtime requirement is Node.js 20 or newer. Run `aegis init` to initialize the installation; the master key is stored in the operating system keychain by default. Add credentials with `aegis vault add`, including a name, service, secret, and permitted domain.

For an unauthenticated local test, start Gate with `aegis gate --no-agent-auth`. For a controlled deployment, create an agent identity, retain the one-time token, grant that agent access to a specific credential, and start Gate with its default authentication behavior.

MCP hosts can use `npx -y @getaegis/cli mcp serve`. The CLI also provides generated configuration commands for Claude Desktop, Cursor, VS Code, Cline, and Windsurf.

## Tools and capabilities

The getaegis/aegis MCP server provides three documented tools:

- `aegis_proxy_request` makes an authenticated API request using a named service and path; the server injects the corresponding credential.
- `aegis_list_services` returns available service names without returning secrets.
- `aegis_health` checks Aegis status.

The MCP implementation applies the Gate pipeline, including domain checks, agent authentication, body inspection, rate limiting, and audit logging. The project also documents examples for Slack, GitHub, Stripe, and OpenClaw, with policies and configuration files for those examples. TLS can be enabled for Gate, and the repository includes security architecture and threat-model documentation for reviewing its trust boundaries and controls.

_Full upstream README: https://allmcps.com/mcp/getaegis-aegis/readme_

