AI Toolkit

Authorization for AI‑native developers.

Install the Vengtoo toolkit in your AI coding agent. Run /vengtoo and your project gets a working authorization layer — SDK installed, middleware wired, policies scaffolded. No context switching, no boilerplate.

Read the docs →Get started free

One command to integrate.

Install the toolkit in Claude Code, GitHub Copilot, or any MCP-compatible agent. Run /vengtoo and it reads your codebase, installs the right SDK for your language, wires middleware into your existing routes, and creates a starter policy file.

Works with Express, Gin, FastAPI and most other popular frameworks. If yours isn't auto-detected, the toolkit prompts you and adapts.

Claude Code / Copilot

# Install the toolkit once
/plugin install vengtoo@vengtoo-toolkit

# In any project
/vengtoo

✔  Detected: Node.js / Express
✔  Installing @vengtoo/sdk...
✔  Wiring authorization middleware...
✔  Writing vengtoo.yaml...
✔  Ready

  Add this to protect a route:
  app.use(vengtoo.authorize())

Everything in the toolkit

Eight skills covering setup, policies, debugging, and MCP wiring.

Command	What it does
/vengtoo	Full setup. Installs SDK, wires authorization middleware, scaffolds policy file. Run this first.
/vengtoo-policies	Reviews your existing policies for gaps, redundancies, or missing actions based on your routes.
/vengtoo-agent	Adds the Vengtoo Agent to your Docker Compose or Kubernetes config and points your SDK at it.
/vengtoo-audit	Queries your decision log and explains recent allow/deny outcomes in plain language.
/vengtoo-check	Runs an authorization check inline — useful for verifying a policy before shipping.
/vengtoo-rbac	Scaffolds a role-based access control model for a resource type you describe.
/vengtoo-abac	Scaffolds an attribute-based access control policy using your entity attributes.
/vengtoo-mcp	Wires the Vengtoo MCP server into your agent configuration for tool-level gating.

MCP

Gate what your AI agents can do.

The Vengtoo MCP server exposes an authorization tool that any MCP-compatible agent can call before taking an action. Run /vengtoo-mcp to wire it automatically, or add it manually to your MCP config.

Every tool call the agent makes is subject to your Vengtoo policies. Deny a tool, and the agent can't use it — even if it tries.

mcp.json

{
  "mcpServers": {
    "vengtoo": {
      "command": "vengtoo-mcp",
      "env": {
        "VENGTOO_API_KEY": "${VENGTOO_API_KEY}",
        "VENGTOO_APP_ID":  "${VENGTOO_APP_ID}"
      }
    }
  }
}