# Let AI agents use your app

> Your AI reads your app's code and makes it easy for AI agents to use: an llms.txt and docs page, a remote MCP server with sign-in and limits, a small CLI, your name and icon in agent apps, and a listing in the MCP Registry. Then it tests the result with Claude, ChatGPT, Cursor and Grok. These are the steps Francisco used to build CloneKit.

- Kit: `@franguzmanx/agent-ready-app` · v1.0.0
- Creator: @franguzmanx (Francisco)
- Type: Delivery · Domain: Engineering · Language: en · Your AI runs it
- Price: Free · License: Own use · No refunds
- Page: https://clonekit.ai/@franguzmanx/agent-ready-app
- Changelog: https://clonekit.ai/@franguzmanx/agent-ready-app/changelog.md

## Deliverables
- AGENT-READY.md: the first 3 to 5 jobs agents can do, what each one changes, and the gaps
- a /llms.txt file and a docs page written for people and agents
- API errors with one shape: code, message and fix
- a remote MCP server (Streamable HTTP) with your first tools, honest annotations and errors that say what to do next
- sign-in for agents (one-tap OAuth, API keys, or both) and rate limits
- a small CLI with login, whoami, setup and --json
- server.json, a server card and the .well-known discovery files, with your name and icon
- Add to Cursor and Add to VS Code buttons, plus connect steps for Claude, ChatGPT, Grok, Claude Code and Codex
- a server the Claude and ChatGPT directories accept (read-only if your app moves money), with a test that keeps it that way
- an entry in the official MCP Registry, and a submission packet for each other directory
- a feedback tool, and a test report from at least three real agents

## Files
- `README.md`: Let AI agents use your app
- `reference/auth-and-limits.md`: Sign-in and limits: how CloneKit does it
- `reference/cli.md`: CLI: how CloneKit does it
- `reference/directories.md`: Directories: how CloneKit did it
- `reference/directory-server.md`: The read-only server for directories: how CloneKit does it
- `reference/discovery.md`: Discovery and branding: how CloneKit does it
- `reference/mcp-server.md`: MCP server: how CloneKit does it
- `reference/testing-with-agents.md`: Testing with real agents: how CloneKit does it
- `templates/agent-ready-plan.md`: AGENT-READY.md
- `templates/directory-packet.md`: <Directory name>: submission packet

## Tools
- MCP Inspector (https://modelcontextprotocol.io/docs/tools/inspector) · est. free
- add-mcp (https://www.npmjs.com/package/add-mcp) · optional · est. free
- MCP Registry (https://registry.modelcontextprotocol.io) · est. free

## Uses
- Model Context Protocol specification (source, https://modelcontextprotocol.io/specification/2026-07-28)
- MCP authorization (source, https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization)
- Writing effective tools for agents (source, https://www.anthropic.com/engineering/writing-tools-for-agents)
- Tool annotations, from the MCP maintainers (source, https://blog.modelcontextprotocol.io/posts/2026-03-16-tool-annotations/)
- The llms.txt format (source, https://llmstxt.org)
- MCP Server Card (source, https://github.com/modelcontextprotocol/experimental-ext-server-card)
- MCP Registry domain authentication (source, https://modelcontextprotocol.io/registry/authentication)
- Cursor MCP install links (source, https://cursor.com/docs/mcp/install-links)
- Claude connector submission (source, https://claude.com/docs/connectors/building/submission)
- Claude connector review criteria (source, https://claude.com/docs/connectors/building/review-criteria)
- ChatGPT app submission (source, https://developers.openai.com/plugins/deploy/submission)
- ChatGPT app guidelines (source, https://developers.openai.com/plugins/app-guidelines)
- Smithery (source, https://smithery.ai/new)
- Glama MCP servers (source, https://glama.ai/mcp/servers)
- CloneKit's llms.txt (template, https://clonekit.ai/llms.txt)
- CloneKit's docs page (template, https://clonekit.ai/docs)

## What you provide
- the app's code, in a repo your AI can read and change
- the app's domain, and access to deploy it
- admin access to the sign-in provider, for one-tap sign-in

## How it runs
1. **Understand:** Researches your business first, then asks you 6 questions it can't answer alone
2. **Choose:** 1 A/B choices with examples
3. **Build:** 9 steps, risky ones ask you first
4. **Deliver:** 6 checks before delivery

## Guardrails
- Ask the person before you deploy, change DNS, create OAuth apps, publish to the MCP Registry or submit to a directory.
- Keep tokens, API keys, signing keys and OAuth secrets in environment variables or the person's secret store. Never put them in code, logs, chat or a test fixture.
- Build on the app's own auth, API and functions. Don't add a framework or a service when the current one does the job.
- A tool that moves money or deletes data keeps the app's own checks and limits on the server.
- Report only what you saw work. A listing, a sign-in or a test you didn't see pass is still open.

## What's new
Agents always run the latest version.

- **v1.0.0** (2026-10-18): First version: nine steps, from the audit to tests with real agents, with code from CloneKit.

## Run it
Connect the `clonekit` MCP (https://clonekit.ai/mcp) and call `start_run("@franguzmanx/agent-ready-app")`.