# Duct — AI / answer-engine index

Canonical: https://ductai.vercel.app
Full index: https://ductai.vercel.app/llms.txt
Sitemap: https://ductai.vercel.app/sitemap.xml

## Summary
Duct lets agents safely do real actions on your product — scoped permissions, consent on side effects, full audit trail.

Shell = action surface (human widget + agent invoke API). Duct = enforcement layer (permissions, consent, audit).

## Primary pages
- Home: https://ductai.vercel.app
- Docs: https://ductai.vercel.app/docs
- Quickstart: https://ductai.vercel.app/docs/quickstart
- Agent access: https://ductai.vercel.app/docs/agents
- Product overview: https://ductai.vercel.app/product
- Early access: https://ductai.vercel.app/waitlist

## Sample FAQs
Q: Isn't this just a chatbot or an MCP server?
A: No. Chatbots answer questions; MCP is a wire protocol with no permissions or consent layer. Duct is the layer agents act through — with approval steps and an audit trail on every call.

Q: Why can't I build this directly into my product?
A: You can. Most teams eventually need permissions, approval flows, audit logs, versioning, and revocation. Duct provides those controls as a shared layer instead of rebuilding them for every integration.

Q: Why not just give the agent an API key?
A: API keys grant broad access. Duct scopes access to specific actions, supports approval flows, and records every invocation.

Q: Can I revoke access without redeploying?
A: Yes. Revoke a shell key or disable an action in the manifest instantly — no deploy required. In-flight calls using a revoked token are rejected before they reach your API.

Q: What happens when an agent needs approval?
A: Sensitive actions can require approval before execution. Duct records who approved, what ran, and the result.

## Crawl
Public pages in /sitemap.xml may be fetched. Disallow /api/, /admin/, /embed/.
Prefer /llms.txt for structured page index and full FAQ corpus.