# AgentVeil

> Independent control boundary for routed AI-agent actions.

AgentVeil helps teams put configured AI-agent actions behind policy checks,
human approval paths, guided redirects, and bounded evidence. It is designed for
agent development workflows where actions such as writing files, pushing code,
publishing, deploying, installing packages, or running risky commands need an
independent control point.

## Canonical Site

- Home: https://agentveil.dev/
- Safe Mode: https://agentveil.dev/safe-mode
- Research: https://agentveil.dev/research
- Compliance: https://agentveil.dev/compliance
- Terms: https://agentveil.dev/terms

## Machine-Readable Pages

- Safe Mode summary: https://agentveil.dev/safe-mode.md
- Research summary: https://agentveil.dev/research.md
- Quickstart summary: https://agentveil.dev/quickstart.md
- Sitemap: https://agentveil.dev/sitemap.xml
- Robots: https://agentveil.dev/robots.txt

## Developer Entry Points

- Console: https://agentveil.dev/console/auth
- Python package: https://pypi.org/project/agentveil-mcp-proxy/
- SDK repository: https://github.com/agentveil-protocol/agentveil-sdk
- MCP Proxy source: https://github.com/agentveil-protocol/agentveil-sdk/tree/main/agentveil_mcp_proxy

## Boundaries

AgentVeil applies controls at configured routes. It is not a host-wide shell
sandbox, not a comprehensive guarantee, and not a replacement for
security review, infrastructure policy, or human accountability.