AI能讀懂你的SaaS嗎？為何llms.txt優於MCP，造福開發者

Sign up

Sign in

Sign up

Sign in

Can AI Read Your SaaS?

--

Listen

Share

What Interface Should SaaS Provide in the Age of AI?

“How do I integrate this API?”

When building side projects, you face this situation multiple times a day. Payment systems, push notifications, deep links, analytics tools… You spend hours reading each SaaS documentation, finding example code, and modifying it to fit your project.

But things are different now.

When you tell Claude or ChatGPT “Write code to create a deep link with Limelink”? Working code appears in 10 seconds. Or at least, it should.

The problem is that most SaaS aren’t ready for the AI era yet.

1. The Era of AI Agents Writing Code

In 2025, the development paradigm has shifted.

AI coding agents like Cursor, Claude Code, and GitHub Copilot have taken over developers’ daily lives. They go beyond simple code autocompletion to generate ”complete code that integrates a specific SaaS into my project.”

But there’s a condition.

The AI must “know” that SaaS.

2. MCP vs llms.txt: Fish vs Fishing Rod

There are two main approaches to building AI-friendly SaaS.

MCP Server Approach: “I’ll do it for you”

MCP (Model Context Protocol) is a protocol that allows AI agents to communicate directly with external services. When a SaaS provides an MCP server, AI agents perform tasks in real-time through that server.

Pros:

Cons:

llms.txt Approach: “I’ll teach you how to do it yourself”

llms.txt is a standard that provides service documentation in a format AI can understand. AI agents read this document and generate code tailored to the user’s project.

Pros:

The Key Difference:

What do developers building side projects really need?

Obviously the latter. Creating one deep link with MCP is easy, but integrating deep link logic throughout your app requires code.

3. llms.txt Usage Examples by Agent

Let’s see how AI agents actually use llms.txt.

Claude (Web/Desktop)

Claude uses llms.txt through the Projects feature.

Claude references the API specs and SDK documentation from llms.txt to generate ready-to-use Swift code.

ChatGPT

ChatGPT uses it directly in GPTs or conversations.

ChatGPT generates platform-specific code based on the documentation.

Cursor / Claude Code

IDE-integrated agents are even more powerful.

Then request from the agent:

The agent:

One enter to integrate SaaS.

4. Interfaces SaaS Must Provide in the AI Era

So what should SaaS providers prepare?

1. Provide llms.txt (Required)

Place an /llms.txt file at your website root. This file contains:

All written in Markdown format. It’s the format AI understands best.

2. Structured API Documentation

3. Platform-specific SDK Examples

Provide example code for major platforms like iOS, Android, React Native, and Flutter. AI references these to generate code matching the user’s environment.

4. (Optional) MCP Server

If llms.txt is “teaching how to fish,” MCP server is a tool for “quick experience.”

Conclusion & CTA

In the AI era, SaaS must be in a state where “when a developer asks, AI can write the code.

If MCP server catches fish for you, llms.txt teaches you how to fish.

What developers building side projects need isn’t one-time help. It’s code that can be permanently integrated into their project.

A SaaS that provides llms.txt is essentially saying to developers:

“Just ask Claude. You’ll get working code instantly.”

One enter to integrate any SaaS.

This will become the standard for developer experience in the AI era.

Start right now:

10 seconds later, working code will be right in front of you.

References

--

--

Written by Limelink

https://limelink.org/ The most easily implementable Dynamic Links

No responses yet

Help

Status

About

Careers

Press

Blog

Privacy

Rules

Terms

Text to speech