cubox-mcp
Enables interaction with Cubox through its CLI, allowing agents to manage bookmarks, highlights, and annotations via a single pass-through tool.
README
cubox-mcp-cli
A generic pass-through MCP server for the official cubox-cli. It exposes exactly one tool, cubox_cli, whose canonical parameter is args: string[] — the argv you would type after cubox-cli on the command line. It also tolerates clients that serialize the array as a string or wrap it in one or more { "item": ... } objects. When Cubox adds, renames, or removes commands, this wrapper does not need to change.
What is this?
This is a Model Context Protocol (MCP) server that wraps the official Cubox CLI. Instead of hard-coding Cubox's command surface, it forwards the argument array straight to cubox-cli and returns the output. It also includes a shallow, fail-safe guard against destructive operations (delete/remove/rm without a force/confirm flag).
Why fork/use this?
- Future-proof:
cubox-clican change its commands and flags anytime; this server stays the same. - One tool for everything:
cubox_cli({ args: ["--help"] })works for any command. - No npm account required: run directly from GitHub with
npx. - Works with any MCP client: tested with mcphub, Claude Desktop, and the MCP Inspector.
- Optional container: ships with a
Dockerfileand a Docker Compose template for HTTP transport.
Quick start
1. Get a Cubox API token
The Cubox backend uses a static API Extension link rather than OAuth, so you log in once and copy the link:
npx -y cubox-cli auth login
# → open https://cubox.pro/web/settings/extensions, enable the API Extension,
# copy the unique link, and paste it back at the prompt
This writes ~/.config/cubox-cli/config.json on the machine where you run it. If your MCP client runs elsewhere, you will need to pass CUBOX_TOKEN as an environment variable (see below).
2. Add to your MCP client
No clone or local install is needed for the host running the MCP client. npx will pull the code from GitHub and install dependencies automatically.
mcphub mcp_settings.json:
{
"mcpServers": {
"cubox": {
"command": "npx",
"args": ["-y", "github:Lascivea/cubox-mcp-cli"],
"env": {
"CUBOX_TOKEN": "your_token",
"CUBOX_SERVER": "cubox.pro"
}
}
}
}
Claude Desktop / other clients: use the same npx command and environment variables.
3. Test it
Call the tool with:
{ "args": ["--help"] }
For compatibility with some function-calling clients, these equivalent forms are also accepted:
{ "args": "--help" }
{ "args": { "item": ["--help"] } }
{ "args": { "item": { "item": ["card", "list"] } } }
The canonical form remains an array of strings. The wrapper unwraps the compatibility forms before applying safety checks and invoking cubox-cli.
Or try a read-only command like:
{ "args": ["folder", "list"] }
Configuration example
| Variable | Required | Default | Notes |
|---|---|---|---|
CUBOX_TOKEN |
yes* | — | The last segment of your Cubox API Extension link (e.g. abcd12345). Required when the MCP host is a different machine than the one where you ran cubox-cli auth login. |
CUBOX_SERVER |
no | cubox.pro |
Cubox server domain. Use cubox.cc for the international instance. |
CUBOX_CLI_BIN |
no | auto | Override the cubox-cli binary path. By default the wrapper resolves node_modules/.bin/cubox-cli, then falls back to PATH. |
MCP_TRANSPORT |
no | stdio |
stdio for mcphub / Claude Desktop / local clients. http for the standalone Docker container. |
PORT |
no | 3000 |
Only used when MCP_TRANSPORT=http. |
No credentials are baked into the image or written to disk by this wrapper.
CUBOX_TOKENandCUBOX_SERVERare read bycubox-cliitself, not by this server.
Docker Compose deployment
Copy the template and edit the environment variables:
cp docker-compose.example.yaml docker-compose.yaml
# edit docker-compose.yaml, set CUBOX_TOKEN
docker compose up -d
The container runs the MCP server over HTTP on port 3000. Register it in your MCP client with a url entry pointing at http://localhost:3000/mcp.
Local development
cd cubox-mcp-cli
npm install
node test-client.js # regression tests against fake-cubox-cli.sh
npm run inspector # interactive MCP Inspector
Design notes
Why only one tool?
A single { args: string[] } schema is the canonical interface and keeps command discovery simple. At runtime, a preprocessing layer also accepts the common array-serialization quirks produced by some function-calling clients: a string, { item: ... }, or nested { item: ... } wrappers. Agents can discover commands live by calling cubox_cli({ args: ["--help"] }) or cubox_cli({ args: ["card", "--help"] }).
What is hard-coded?
- The global
-o jsonflag, with a fallback to plain text if a sub-command rejects it. - A shallow guard against destructive operations: any call containing
delete/remove/rmmust also contain--force,--yes,-y, or--confirm. If Cubox renames its delete command, this guard simply stops matching — it never blocks something new incorrectly.
Upgrading cubox-cli later
Nothing to do. npx resolves the dependency fresh from package.json whenever the cache is rebuilt. To force a refresh: npx clear-npx-cache or bump the cubox-cli version range in package.json.
Companion projects / related tools
cubox-cli— the official Cubox CLI this server wraps.- mcphub — an MCP server manager that pairs well with this project.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.