Skip to main content
Glama
sam-14uel

DeskLink

by sam-14uel
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Local

DeskLink

Turn your local computer into a remotely accessible MCP endpoint for cloud AI agents.

DeskLink runs a secure local filesystem MCP server (Streamable HTTP) and exposes it through a temporary public tunnel (ngrok by default) so tools like Grok, Claude, or Cursor can access permitted directories on your machine.

Quick start

pip install desklink
# Optional: export NGROK_AUTHTOKEN=your_token   # required for free ngrok
desklink start

Example output:

DeskLink v0.1.0

✓ Filesystem MCP started
✓ Local MCP: http://127.0.0.1:8765/mcp
✓ Tunnel connected

Remote MCP endpoint:
https://abc123.ngrok-free.app/mcp

Auth token: sk-desklink-...
Include as Authorization: Bearer <token> when connecting.

Status: ● Connected
Tools: 5
Allowed dirs: 1

Press Ctrl+C to stop.

Connect the remote URL + bearer token from your AI client.

Related MCP server: Remote MCP Hub

Installation

pip install desklink
# or for development
uv pip install -e ".[dev]"

Requires Python 3.10+.

Configuration

DeskLink uses sensible defaults and can be overridden via environment variables (DESKLINK_*) or CLI flags.

Default allowed directory: ~/Documents.

desklink start --allow ~/projects --allow /tmp/workspace --read-only
desklink start --port 9000 --tunnel none   # local only

Persistent token is stored at ~/.desklink/token (mode 0600).

Security

  • Bearer token authentication is enabled by default.

  • Filesystem access is restricted to explicitly allowed directories.

  • Relative paths are resolved against the first allowed directory.

  • Path traversal outside allowed roots is rejected.

  • The public tunnel URL alone does not grant access; the token is required.

Future versions will support JWT, machine approval, organization tokens, and a cloud relay.

Architecture

Modular by design:

  • cli – Typer CLI

  • server – HTTP + lifecycle

  • filesystem – restricted filesystem tools

  • tunnel – pluggable providers (ngrok, none, future Cloudflare / DeskLink relay)

  • auth – token verification (extensible)

  • config / models – typed configuration

Additional MCP servers (terminal, browser, git, docker, etc.) can be added later and composed behind the same endpoint.

Development

uv venv
source .venv/bin/activate
uv pip install -e ".[dev]"
pytest

License

MIT

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • LanternFS

    deviprasadshetty-dev
    -
    license
    Not graded
    quality
    B
    maintenance
    Enables MCP clients like ChatGPT to securely access and manage a controlled local directory over a remote, OAuth-protected endpoint, with configurable filesystem roots and ngrok tunneling.
    -

  • Remote MCP Hub

    etern1ty-crypto
    A
    license
    Not graded
    quality
    C
    maintenance
    Enables AI assistants to securely access and control a Windows machine remotely, running commands and managing files via an MCP server protected by Bearer-token authentication and exposed through a Cloudflare Tunnel.
    MIT

  • Remote MCP Hub

    etern1ty-crypto
    A
    license
    Not graded
    quality
    C
    maintenance
    Enables remote access to local filesystem and Windows command execution through MCP-compatible clients like Notion AI, Claude, and Cursor via secure public tunnels with bearer authentication and proxy support.
    MIT

  • Aura

    XRSec
    A
    license
    Not graded
    quality
    A
    maintenance
    Enables AI assistants to securely access local machine capabilities such as file operations, shell commands, Skills, and Pi Tools through a controlled MCP interface with token authorization and multiple tunnel options.
    3 npm
    1
    MIT