How does it detect code blocks?

It scans for standard markdown fences — three or more backticks, an optional language tag on the opening line, the code body, then a matching closing fence. Each block keeps its language label so you can tell a Python snippet from a shell command at a glance.

Does it handle multiple languages in one response?

Yes. A reply that mixes a TypeScript file, a shell command, and a JSON config produces three separate labeled blocks. Blocks without a language tag are labeled as plain text.

Can it handle nested or indented fences?

It handles the common case of top-level triple-backtick fences reliably. Deeply nested fences inside other fences are rare in LLM output and may not split perfectly — in practice almost every model reply uses flat, top-level fences.

Is my code uploaded anywhere?

No. The extraction runs entirely in your browser. Nothing is sent to a server or stored.

What is the Code Block Extractor?

Parse markdown-formatted LLM output and extract every fenced code block, grouped and labeled by language tag. Copy any block individually or copy all at once — fast, local, and no setup. It runs free in your browser on Gera Tools, with nothing uploaded.

Code Block Extractor

Name: Code Block Extractor
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

Get one useful tool a week

Like this tool? Enter your email and we'll send you one genuinely useful Gera tool a week — plus a link to come back to this one. No spam, one-click unsubscribe any time.

Code block extractor

LLMs love to wrap code in markdown fences and surround it with explanation. When you just want the code — maybe several files across several languages in one reply — copying it by hand is tedious and error-prone. This tool reads the whole response, pulls out every fenced code block, labels each by language, and lets you copy any one of them, or all of them at once.

How it works

The extractor scans for standard markdown fences: an opening line of three or more backticks with an optional language tag, the code body, and a matching closing fence. It captures each body verbatim and records the language from the opening line — python, ts, bash, and so on — defaulting to “text” when no tag is present. Blocks are returned in document order, each shown with its language label and an individual copy button, plus a copy-all action that joins them for a quick paste into your editor.

When to use this tool

The most common scenario is a “generate a full project scaffold” prompt. A single ChatGPT or Claude reply can include a package.json, a TypeScript file, a Dockerfile, and a shell install script — all wrapped in separate fences. Rather than scrolling through the prose and manually selecting each block, paste the entire response here and grab every file in seconds.

Other useful cases:

Code review replies — models often return the fixed code alongside commentary; extract just the corrected file.
SQL generation — pull the query block away from the explanation text before running it in a client.
Multi-step tutorials — extract every command block into one list so you can run them in order without re-reading the tutorial.

Worked example

Suppose you paste a model response that contains two TypeScript snippets separated by explanation: one for a greet function and one for its test. The tool finds both fenced blocks, labels each as ts, and gives each its own copy button. The copy-all action concatenates them so you can paste both into your editor at once — two files in one paste, no manual hunting through prose.

Notes and tips

Language labels come from the fence. Whatever tag the model wrote after the opening backticks is what you see — accurate as long as the model tagged it.
Untagged blocks become “text”. Plain triple-backtick blocks still extract; they are just labeled generically.
Flat fences work best. Top-level fences — the format virtually every LLM uses — extract cleanly; rare nested fences may not split perfectly.
No stripping prose. The tool preserves each block verbatim, including leading or trailing blank lines inside the fence, so indentation in YAML/Python/etc. is kept intact.
Stays in your browser. Nothing leaves the page, so paste proprietary code without concern.