CCM
/MCP
SkillsMCPMarketplacesDigestToolsAdvertise

This week in Claude

Every Monday: Claude Code, Agent SDK, MCP, and the Anthropic platform moves worth your time.

Skills by Category
Frontend DevelopmentBackend & APIsTesting & QASecurityDevOps & CI/CDGit & Pull RequestsDocumentationCode Review & QualityAI & Agent BuildingSkill Development
MCP Servers by Category
Sales & MarketingWeb & Browser AutomationDatabasesAI & LLM ToolsCloud & InfrastructureCommunication & MessagingDeveloper ToolsDesign & CreativeDocuments & KnowledgeSearch & Web Crawling
Marketplaces by Category
AI Agents & OrchestrationLLM IntegrationDevelopment ToolsFrontend & UIBackend & APIsDatabasesTesting & Code QualityDevOps & CloudSecurity & ComplianceGit & Version Control

Claude Code Marketplaces

Discover Claude Code plugins, extensions, and tools. Automatically updated directory of Anthropic Claude AI marketplaces with development tools, productivity plugins, and integrations.

Resources

  • Browse Skills
  • Browse MCP Servers
  • Browse Marketplaces
  • Skill index
  • MCP index
  • Marketplace index
  • Plugins Reference

Community

  • About
  • Tools
  • Feedback
  • Privacy Policy
  • Advertise

Built for the Claude Code community with Claude Code by mertbuilds.com

Independent project, not affiliated with Anthropic
ridget avatar

Supernote

ridget/supernote-mcp
STDIOregistry active
Summary

Gives your AI direct access to a Ratta Supernote e-ink tablet over Wi-Fi. Captures the live screen while you're sketching or handwriting, reads saved notebooks as recognized text or rendered page images, browses the file tree, and uploads PDFs or documents for annotation. Built on reverse-engineered LAN APIs (screen mirroring on port 8080, Browse & Access file server on 8089), so it's unofficial and could break with firmware updates. Useful when you want to pull handwritten notes straight into a conversation, have the agent summarize meeting scribbles, or push reference material to the device without leaving your terminal. Read-only except for uploads, with automatic network discovery if your tablet's DHCP lease changes.

CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
MCP-ready Email SendingMCP-ready Email Sending
MCP-ready Email Sending
Plug Mailtrap into your AI workflow and let it handle the email.
Connect Mailtrap MCP →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Capacitor - Shared memory for your team’s coding agents.
Capacitor - Shared memory for your team’s coding agents.
Make coding agent sessions - Searchable, Shareable, Vendor-neutral & Scored.
Try For Free →
CodeScene MCP ServerCodeScene MCP Server
CodeScene MCP Server
Your agent targets a perfect 10 Code Health score. Deterministic. Every commit.
Try For Free →
Give your AI the whole web as clean markdownGive your AI the whole web as clean markdown
Give your AI the whole web as clean markdown
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
belt - the only tool your agent needs
belt - the only tool your agent needs
belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →
inference shell
inference shell
create and run specialised agents in minutes
build now →
CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
MCP-ready Email SendingMCP-ready Email Sending
MCP-ready Email Sending
Plug Mailtrap into your AI workflow and let it handle the email.
Connect Mailtrap MCP →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Capacitor - Shared memory for your team’s coding agents.
Capacitor - Shared memory for your team’s coding agents.
Make coding agent sessions - Searchable, Shareable, Vendor-neutral & Scored.
Try For Free →
CodeScene MCP ServerCodeScene MCP Server
CodeScene MCP Server
Your agent targets a perfect 10 Code Health score. Deterministic. Every commit.
Try For Free →
Give your AI the whole web as clean markdownGive your AI the whole web as clean markdown
Give your AI the whole web as clean markdown
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
belt - the only tool your agent needs
belt - the only tool your agent needs
belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →
inference shell
inference shell
create and run specialised agents in minutes
build now →

supernote-mcp

An MCP server for the Ratta Supernote e-ink tablet. It lets an AI agent see and work with your Supernote over the local Wi-Fi network:

  • Grab the live screen while you sketch or handwrite — "what did I just draw?"
  • Browse and read your saved notebooks — recognized handwriting as text, or pages as images.
  • Upload files to the device for you to read or annotate.

The idea is collaborative prompting: handwrite on the tablet during a planning session and pull it straight into the conversation, or have the agent read and reason over notes you saved earlier.

Every tool is lazy — it touches the device only when invoked, so a session that never calls one makes no network calls to the tablet.

[!IMPORTANT] Unofficial, community-built project — not affiliated with, authorised by, or endorsed by Ratta / Supernote. "Supernote" and "Ratta" are trademarks of their respective owner, used here only for identification. It relies on reverse-engineered, undocumented LAN interfaces (screen mirror + Browse & Access) that may change or break with firmware updates. Provided as-is under the MIT license — use at your own risk.

Built on supernote-typescript for screen capture and .note parsing.

Prerequisites

  • A Supernote that supports Screen Mirroring (Manta / Nomad / A5X / A6X-class devices).
  • The tablet and the machine running this server on the same Wi-Fi network, with no VPN or proxy active — either will break the LAN mirror connection.
  • For the npx install path: Node.js ≥ 18. For bunx: Bun. The standalone binary needs neither.

Enable mirroring and find the IP

  1. On the Supernote, open the sidebar and turn on Screen Mirroring (a.k.a. "Cast"/"Screencast").
  2. A popup shows an address like 192.168.1.42. That's the device IP. The mirror serves on port 8080 (http://<ip>:8080/screencast.mjpeg); the server appends :8080 for you.
  3. Keep mirroring on while you want the agent to be able to snapshot the canvas.

Set the IP once via the SUPERNOTE_IP environment variable, or pass it per-call as the tool's ip argument (an explicit argument wins over the env var).

Browse & Access (for the file tools)

supernote_list_files (and the note tools) use a separate device feature from screen mirroring: Browse & Access, the built-in Wi-Fi file server on port 8089.

  1. On the Supernote, swipe down from the top of the screen to open the drop-down toolbar and tap Browse & Access.
  2. A popup shows an address like http://192.168.1.42:8089 and stays open while it's active. The device IP is the same as for mirroring; the tools target port 8089 automatically.
  3. Keep the popup open while you want the agent to read or upload files.

Browse & Access and Screen Mirroring are independent toggles — enable whichever the task needs (or both). Discovery probes each port, so the device can be found with only one of them on.

Coping with a changing IP

The mirror address is a DHCP lease and can change between sessions. Two ways to stay robust:

  • Pin it (recommended): add a DHCP reservation on your router so the tablet always gets the same IP, then SUPERNOTE_IP never goes stale.
  • Automatic fallback: if the configured address is unreachable — or none is set — the server scans the local network for the mirror, uses the device it finds, and logs the new IP so you can update SUPERNOTE_IP. The scan only runs as a fallback (never on the fast path) and probes port 8080 across your subnet for the mirror's multipart stream. Disable it with SUPERNOTE_DISCOVER=0 if you'd rather it fail fast than scan the LAN.

Install & register with Claude Code

Pick whichever distribution suits you — none require this repo to be checked out.

# Recommended: published npm package via Node's npx
claude mcp add supernote --scope user --env SUPERNOTE_IP=192.168.1.42 -- npx -y supernote-mcp

# If you already have Bun
claude mcp add supernote --scope user --env SUPERNOTE_IP=192.168.1.42 -- bunx supernote-mcp

# Standalone binary (no runtime needed) — download the asset for your platform from the
# latest GitHub Release (supernote-mcp-darwin-arm64, -darwin-x64, -linux-x64, -linux-arm64,
# -windows-x64.exe), make it executable, then register its path:
chmod +x ./supernote-mcp-darwin-arm64
claude mcp add supernote --scope user --env SUPERNOTE_IP=192.168.1.42 -- /path/to/supernote-mcp-darwin-arm64

Restart Claude Code, then try: "Snapshot my Supernote and tell me what I drew."

On macOS the downloaded binary is unsigned; the first run may be blocked by Gatekeeper. Right-click → Open once, or run xattr -d com.apple.quarantine /path/to/supernote-mcp-darwin-arm64.

The tools

All tools accept an optional ip (device IP, optionally :port; defaults to SUPERNOTE_IP, else a LAN scan) and return a clear, actionable message on failure rather than hanging.

ToolNeedsInputReturns
supernote_snapshotScreen Mirroring (8080)ip?the live screen as image/png
supernote_list_filesBrowse & Access (8089)ip?, path?a listing — name, folder?, size, date, and a path to pass on
supernote_read_noteBrowse & Access (8089)ip?, patha note's recognized handwriting/text per page (or a note that recognition hasn't run)
supernote_render_noteBrowse & Access (8089)ip?, path, pages?note pages as image/png (all pages, capped at 20, or the pages you pick)
supernote_upload_fileBrowse & Access (8089)ip?, path, directory?, filename?uploads a local file to the device (the only tool that writes to it)

Failures point at the usual causes — wrong IP, the relevant feature turned off, or the host/device not sharing a VPN-free Wi-Fi network — and time out fast (10s) rather than hanging.

All tools except supernote_upload_file are read-only. supernote_upload_file writes a file to the device (it reads a local file the server can access and POSTs it over Browse & Access).

Choosing the right tool

The server also ships this routing as MCP instructions, so a connected agent picks the right tool on its own. The map:

You want…Say something likeTool
What's on the screen right now"what did I just draw?", "look at my screen"supernote_snapshot
To find a saved note"what notes do I have?", "find my note about X"supernote_list_files
To read a note's words"read / summarise my meeting notes"supernote_read_note
To see a note's pages"show me that sketch", "look at page 2"supernote_render_note
To put a file on the tablet"send this PDF to my Supernote"supernote_upload_file

Saved-note flow: supernote_list_files → take the entry's path → supernote_read_note (text, cheap — prefer for words) or supernote_render_note (images, for drawings or notes without recognized text). supernote_snapshot is separate — it's the current screen, not a saved file.

Local development

Bun is used only to run, typecheck, and build — it is not required by end users. This repo ships an optional devbox environment that provides Bun reproducibly; you can equally use a system Bun install.

bun install

# Capture-first verification against a real device (writes a PNG you can open):
bun run src/capture.ts --ip 192.168.1.42 --out frame.png

# Browse & Access (port 8089) — list a directory, download a file, upload a file:
bun run src/browse.ts list --ip 192.168.1.42 --path /Note
bun run src/browse.ts get  --ip 192.168.1.42 --path /Note/obsidian/x.note --out x.note
bun run src/browse.ts put  --ip 192.168.1.42 --path ./doc.pdf --dir /INBOX

# Run the MCP server over stdio:
bun run src/server.ts

# Inspect the tool interactively:
bunx @modelcontextprotocol/inspector bun run src/server.ts

bun run test           # unit tests (bun:test)
bun run typecheck      # tsc --noEmit
bun run build          # node-compatible bundle -> dist/server.js (the npx/bunx entry)
bun run build:binary   # self-contained executable -> dist/supernote-mcp

bun build --compile cross-compiles too, e.g. bun build --compile --target=bun-linux-x64 src/server.ts --outfile dist/supernote-mcp-linux-x64.

How it works

src/capture.ts wraps fetchMirrorFrame(${ip}:8080), which parses the device's multipart/x-mixed-replace stream, extracts a frame, and decodes it via image-js. The frame is re-encoded to PNG (sidestepping PNG-vs-JPEG part-encoding differences across firmwares) and returned as base64. src/browse.ts is a client for the Browse & Access HTTP file server (port 8089): it lists directories (parsing the listing page's embedded JSON) and downloads files. src/resolve.ts holds the shared address resolution + LAN-scan fallback used by both, and src/discover.ts does the subnet scan (probing the mirror or Browse & Access endpoint to recognise the device). src/server.ts registers the tools on an McpServer over the stdio transport.

License

MIT

Featured
CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
MCP-ready Email SendingMCP-ready Email Sending
MCP-ready Email Sending
Plug Mailtrap into your AI workflow and let it handle the email.
Connect Mailtrap MCP →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Capacitor - Shared memory for your team’s coding agents.
Capacitor - Shared memory for your team’s coding agents.
Make coding agent sessions - Searchable, Shareable, Vendor-neutral & Scored.
Try For Free →
CodeScene MCP ServerCodeScene MCP Server
CodeScene MCP Server
Your agent targets a perfect 10 Code Health score. Deterministic. Every commit.
Try For Free →
Give your AI the whole web as clean markdownGive your AI the whole web as clean markdown
Give your AI the whole web as clean markdown
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
belt - the only tool your agent needs
belt - the only tool your agent needs
belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →
inference shell
inference shell
create and run specialised agents in minutes
build now →

Configuration

SUPERNOTE_IP

Device IP (optionally with :port) from the Supernote mirroring popup. Optional — if unset, the server scans the LAN for the mirror.

SUPERNOTE_DISCOVER

Set to 0 to disable the LAN-scan fallback used when SUPERNOTE_IP is unset or unreachable.

Categories
AI & LLM ToolsDesign & Creative
Registryactive
Packagesupernote-mcp
TransportSTDIO
UpdatedJun 3, 2026
View on GitHub

Related AI & LLM Tools MCP Servers

View all →
rillcoin avatar
RillCoin

rillcoin/rill

AI agent wallets, Proof of Conduct reputation, and concentration decay on L1
rishavdutta-kgp avatar
sentimatix

rishavdutta-kgp/sentimatix

# Sentimatix: Indian Stock Market Intelligence Sentimatix is a high-fidelity financial intelligence API and MCP server designed specifically for the **National Stock Exchange of India (NSE)**. It provides AI agents, algorithmic traders, and quantitative developers with deep, structured insights into over 2,200 Indian equities. ## What it does By combining real-time financial news ingestion with advanced NLP sentiment analysis and historical market data, Sentimatix acts as an automated quantitative research assistant. It empowers LLMs and AI agents to make data-driven decisions regarding the Indian stock market without needing expensive institutional data terminals. ## Key Capabilities - **Entity-Level Sentiment Analysis:** Aggregates and scores sentiment from major Indian financial news sources (Moneycontrol, Economic Times, Mint). - **Deep Stock Research:** Generates comprehensive single-stock and comparative research reports. - **Technical Analysis:** Calculates RSI, MACD, Bollinger Bands, and moving averages on demand. - **RAG Evidence Search:** Provides direct semantic search over a vast, continuously updated financial news corpus. - **Market Movers:** Explains sudden price changes using a fusion of technical metrics and recent news sentiment. ## Perfect For - AI Trading Bots analyzing NSE stocks - Portfolio Management assistants - Retail investors seeking institutional-grade, automated research
rivradev avatar
Recite MCP

rivradev/recite-mcp

Local MCP server for Recite receipt processing with ledger and memory tools.
rjandino avatar
Zopaf

rjandino/zopaf

Negotiation math engine. Pareto frontier, counteroffer generation, zero LLM tokens.
rnwy avatar
RNWY Trust Intelligence

rnwy/mcp

Check if an AI agent is trustworthy. Sybil detection, signed attestations, 150,000+ agents. Free.
robert19001-cmyk avatar
mcp-lint

robert19001-cmyk/mcp-lint

Lint MCP tool schemas for cross-client compatibility and agent preflight safety.