Approximate Filters MCP server

This is a small set of helper programs that let your AI test different ways of checking whether a word is in a list. It is made for people who want to see how Bloom filters, Cuckoo filters and similar tools behave, without writing code themselves. If you have ever wondered how search boxes or blocked-word lists stay fast, this is a hands-on way to look.

What this MCP server does

You ask your AI something like whether a word is in a keyword list. The AI sends that request to this helper program. The helper picks one of the filter structures, such as Bloom or Cuckoo, and runs the check on the data you gave it. The result comes back into the chat as a simple answer, a number, or a small table. You can also ask it to compare filters side by side on the same word list.

Click to zoom

What you can do with it

• Build a filter from a list of words you paste in
• Check whether a word is inside a filter
• Add a new word to a filter
• Delete a word from filters that support deletion
• Ask for a prefix or range query on filters that support it
• See the estimated memory each filter uses
• Measure the false positive rate of a filter

Try asking your AI

• “Build a Bloom filter from these keywords: apple, banana, cherry, date”
• “Is the word grape in the Cuckoo filter”
• “Compare memory usage of Bloom, Counting Bloom and Cuckoo on the same list”
• “What is the false positive rate of the SuRF filter on this dataset”

What it gives back to you

You get short answers in the chat: yes or no for membership checks, a number for memory usage, a percentage for false positive rate, and small lists or tables when you ask for comparisons. If you build a filter, it stays available for follow-up questions in the same session. Nothing is written to your files unless you ask for it.

Before you start

Who it's for

Students, teachers and data-curious office workers who want to see how approximate filters behave without writing code.