floorp_get_content returns unbounded HTML, causing context overflow in LLMs

[johnzfitch]

Problem

The floorp_get_content tool returns the entire HTML of a page without any size limits. Modern web pages can easily exceed 50-100KB of HTML, which:

• Quickly exhausts LLM context windows
• Makes the tool impractical for real-world use
• Can cause billing issues on usage-limited accounts

Example

A simple Kagi search page returned ~25,000 tokens of HTML, immediately filling context.

Environment

• floorp-mcp-server via npx
• Claude Code CLI

Suggestions

There are many ways to try and solve this problem. I'm going to try some stuff out on my end and hope you will too. I'll report back whatever results I have.

[johnzfitch]

@surapunoyousei what do you think about converting it into LLMS.txt format (FastHTML)?