Nuttle MCP server

Nuttle

Everything you browse, saved for you — and one question away for your AI agent. Export Reddit comments with every nested reply · Save X posts and replies · Download the videos you watched · Query it all from Claude Code over MCP

nuttle.dev · Claude Code MCP guide · 中文

As you scroll Reddit, X, YouTube or Instagram, Nuttle quietly keeps the posts, the full comment trees and the original videos on your computer. Export a thread to CSV, download a video, or have your AI agent turn a thousand comments into one feature-request list.

• ✓ No login or API keys ✓ No auto-scrolling, no bulk requests ✓ Data stays on your machine

What it does

• Export Reddit comments — every nested reply included. Open a thread and scroll; one click to CSV.
• Save X posts and their replies — text, metrics, replies and images, searchable.
• Download the videos you watched as original files; Reddit's split audio and video are merged for you.
• A thousand comments, one feature-request list. Hand a scope to your own AI agent (Claude Code, Codex, any MCP client); every quote links back to the original comment.
• Filter and sort · export to CSV · star and tag · back to the source · capture any website with Discover.

Use it with Claude Code or any MCP client

Nuttle runs a local MCP server (streamable HTTP, localhost only) while the app is open:

claude mcp add --transport http nuttle http://127.0.0.1:8787/mcp

Or install the Claude Code plugin — same connection, plus a skill that turns any site you browsed into a reusable mapping card:

claude plugin marketplace add shiyue1250/nuttle-registry
claude plugin install nuttle@nuttle

Claude Desktop (and other clients that only launch local commands) use the nuttle-mcp bridge in claudedesktopconfig.json:

{ "mcpServers": { "nuttle": { "command": "npx", "args": ["-y", "nuttle-mcp"] } } }

Other clients (Cursor, Codex, …) point at http://127.0.0.1:8787/mcp; for Codex add [mcp_servers.nuttle] with url = "http://127.0.0.1:8787/mcp" to ~/.codex/config.toml.

Then ask things like:

• "What did people complain about in the Reddit threads I read today?"
• "Wait while I scroll this thread, then summarize the replies."
• "Group these comments into feature requests, with quotes."
• "Download the videos from those posts."

32 tools cover a post with its full comment tree in one call, waiting for new data while you scroll, raw captures from any site (with code-free field mapping), media downloads, and insight jobs whose quotes are checked against the captured comments. Tags you set yourself can't be overwritten by the agent. Guide: nuttle.dev/en/claude-code-mcp

Supported sites

Reddit · X · YouTube · Instagram · Amazon · Xiaohongshu (RedNote) · WeChat Channels · WeChat articles · Douyin · Kuaishou · Bilibili — plus any other site through Discover.

Get started

1. Download for Windows (about 6 MB, no sign-up), or from nuttle.dev
2. Turn on the sites you want under Plugins
3. Browse as usual — come back to Nuttle and it's all there

FAQ

Will my account get banned? Nuttle sends no extra requests and never scrolls for you; it only keeps what the page already loaded while you browsed, the same as you scrolling yourself. Platform rules are up to each platform, and we can't guarantee them.

Shortened. The full README is on GitHub.