Make your website usable by AI agents.

Descuff analyzes your existing app and exposes its capabilities through agent-ready standards, without changing the experience for humans.

Read why it exists View on GitHub

How it works

Expose what already exists.

Understand

Descuff analyses your application code to map routes, APIs, forms, capabilities, and existing standards.

Plan

It recommends the right agent-facing standards and writes an evidence-backed plan for implementation.

Implement

Use the plan with Codex, Cursor, Claude Code, or another coding agent to add the required files and metadata.

Validate

Descuff verifies routes, standards, security boundaries, and readiness before you ship.

Use it with your coding agent

Install -> Descuff -> validate.

Scan your local Next.js app. Let your assistant implement the plan. Then validate the before and after.

Step 01

Run Descuff in your assistant

Install the workflow for Codex, Claude Code, or Cursor. No dashboard. Your app stays local.

What are you coding with?

Codex

$ npx descuff install --platform codex
> $descuff .   # inside Codex

Descuff currently supports local Next.js App Router and Pages Router projects.

Then star it on GitHub and send feedback from the field.

Step 02

Get the baseline

Descuff scans routes, APIs, forms, auth, Server Actions, and existing standards.

.descuff output

.descuff/
  baseline.json
  model.json
  plan.md
  skill-evidence-packet.json

Focused evidence, not repo-wide guessing.

Step 03

Let your assistant implement the plan

Your coding agent adds safe standards while preserving existing UI and behavior.

validate changes

$ npx descuff enrich .
$ npx descuff finish .

Evidence-backed enrichment
Conservative plan
No hidden LLM call
No sensitive tools without approval

Step 04

See the before and after

Descuff reports readiness, standards added, failures, warnings, and blockers.

The result is validated, not vibes.

example result

Readiness: 60/100 -> 85/100
Standards: none -> llms.txt, Schema.org, OpenAPI
Failures: 0 -> 0
Warnings: 0 -> 0

llms.txt

Public agent-readable site summary.

Schema.org

Structured entities agents and search systems can understand.

OpenAPI

Documented API operations for safe integration.

API Catalog

Discoverable service metadata.

WebMCP

Browser-registered tools for safe public read capabilities.

Feedback

Send notes from the field.

Tell us what worked, what broke, or what would make Descuff easier to use in your own Next.js apps.