Download Claude Artifacts - Save Code & Docs
Download Claude artifacts as files. Save every code block, doc, and app in a conversation with the correct file extension.
- Users
- 5
- Works on
- Chrome
- Version
- 3.0.3
Claude writes you a working script, then leaves you to copy it out of the chat and remember whether it was .py or .ts. A button at the bottom right of the conversation shows how many artifacts and code blocks it contains, and one click writes all of them to disk.
There's no built-in way to download Claude artifacts, so you end up copying each code block, doc, and app out of the chat by hand. This does it in one click.
It scans the open claude.ai conversation for every artifact and inline code block, then saves each one as a real file with the right extension (.py, .jsx, .html, .md, .svg, and so on). Files go into a folder named after the conversation, inside your Downloads. If it can't tell the language, the file saves as .txt so you never lose the content.
Independent extension, not affiliated with Anthropic or Claude.
Screenshots

When you'd reach for it
A conversation that produced a dozen files
A long debugging session leaves code scattered across twenty messages. One click collects every code block and open artifact, names each file, and downloads them into a folder taken from the conversation title. Identical blocks are collapsed, so a snippet Claude repeated four times while iterating is saved once.
Grabbing a single block without the rest
Hover any code block and a download icon appears in its corner. It saves just that block, using the same naming rules as the bulk button. This is the quicker path when you asked for one config file and do not want the rest of the transcript's code on disk.
Files that keep their extension
Extensions come from the block's own language label, and from a filename in a first-line comment when there is one — a block starting with # app/models.py saves as models.py. When the label is missing, content is sniffed for a doctype, an <?xml prolog, an <svg root, or a JSON shape, and anything still unrecognized becomes .txt rather than being dropped.
Permissions & privacy
It runs only the code shipped inside the package — no remote code is fetched.
Questions people ask
How do I download an artifact from Claude?
Open the conversation on claude.ai and click "Download all artifacts" at the bottom right, or hover one code block and click its download icon. Files land in your Downloads folder, inside a subfolder named after the conversation. The count on the button updates as the page changes, so it reflects new code as a reply streams in.
What file extension will my code get?
It maps the language label on the block to an extension, covering roughly ninety names — javascript, tsx, python, rust, kotlin, yaml, sql, mermaid, solidity, and so on. A filename written in a first-line comment wins over the label, and an artifact title that already ends in an extension is used as-is. Files that cannot be identified save as .txt so the content is never lost.
Does it capture the whole artifact from the side panel?
It reads the artifact panel's code editor line by line, supporting both CodeMirror and Monaco. One thing to know: those editors only render the lines currently in view for long documents, so scroll a long artifact to the end before downloading it. Code blocks in the message stream itself are read in full and are not affected by this.
What happens if two files end up with the same name?
Within one download the second copy becomes name-2.ext, the third name-3.ext, and so on. Against files already on disk, Chrome's own uniquify behavior applies, so an earlier download is never overwritten. Blocks whose content matches are treated as one file before naming even starts.
What permissions does it need to read my conversations?
The manifest declares one permission, downloads, used to write each file without a Save As prompt. There are no host permissions; the content script is restricted to https://claude.ai/* by the manifest's match pattern, so it does not run anywhere else. Artifact text is read from the page you have open and passed to Chrome's download API, and the extension does not use chrome.storage.
Related extensions
SEO Toolkit5K+ usersAll-in-one SEO analysis: on-page score, SERP preview, keyword density, broken links, heading structure, and meta tag viewer.
BugSnap2 usersOne-click bug reports with screenshot, console logs & environment info.
AI Credit Meter - v0, Lovable, BoltTrack your remaining credits on v0, Lovable and Bolt, with burn-rate and a projected run-out date. Read-only, all local.
GitHub Unfollowers & Un-stars TrackerSee who unfollowed you on GitHub, and who un-starred your repo — the thing GitHub never tells you. Official API, all local.
Page RulerMeasure any element on a webpage in pixels. Drag to measure areas, hover to see dimensions. Free ruler for designers.
Stack Overflow 한국어 (Korean)Use Stack Overflow every day but find the English interface a bit of friction?