Rustunnel

Rustunnel — Secure Tunnel Management

Expose local services (HTTP / TCP / UDP / P2P) through public URLs using rustunnel. Perfect for testing webhooks, sharing local development, database access, and AI agent workflows. Works through the rustunnel MCP server, so any MCP-capable harness (Claude Code, Claude Desktop, Codex, Cursor, Windsurf, Cline, or a custom agent) can drive it.

When to Use

• Webhook testing — receive webhooks from Stripe/GitHub/etc. on a local server
• Demo sharing — share local development with stakeholders
• CI/CD & previews — expose preview environments
• Database / TCP access — expose PostgreSQL, Redis, SSH, raw TCP
• UDP services — game servers, DNS, custom UDP protocols
• Peer-to-peer — direct, NAT-punched connections between two peers
• Load balancing — spread traffic across multiple local backends with health checks

Authentication — you need an API token (one-time)

rustunnel requires an API token. Get one once, then reuse it:

1. Hosted (rustunnel.com): sign up free at https://rustunnel.com → Dashboard → API Keys → create a key.
2. Self-hosted: create one with rustunnel token create --name agent (needs the server admin token), or from your dashboard.

Where the token lives (in priority order):

1. The RUSTUNNEL_TOKEN environment variable, set once in your MCP client config. This is the recommended setup — when it's set, you do not need to pass token on tool calls and you should not ask the user for it again.
2. A token argument passed explicitly to a tool call.
3. ~/.rustunnel/config.yml (auth_token:) for the CLI.

If no token is configured, ask the user once: "What's your rustunnel API token? Get one free at https://rustunnel.com → Dashboard → API Keys." Then prefer storing it as RUSTUNNEL_TOKEN in the MCP config so it persists.

IMPORTANT: Prefer MCP Tools over the raw CLI

Use the MCP tools for tunnel management — they handle the process lifecycle automatically (no orphaned processes, automatic cleanup, tunnel IDs for tracking). Only fall back to the CLI (get_connection_info) when the agent cannot spawn subprocesses (e.g. a cloud sandbox).

MCP Tools

create_tunnel

Open a tunnel and get a public URL. The token may be omitted when RUSTUNNEL_TOKEN is set.

health_check object: { type: "tcp"|"http", path, interval_secs, timeout_secs, max_failed, expect_2xx, alert_webhook }.

Returns:

jsonCopy code
{ "public_url": "https://abc123.eu.edge.rustunnel.com", "tunnel_id": "a1b2c3d4-...", "protocol": "http" }

Tunnel stays open until close_tunnel is called or the MCP server exits.

close_tunnel

Close a tunnel by ID. The public URL stops working immediately.

list_tunnels

List active tunnels (public URL, protocol, traffic count). token optional¹.

get_tunnel_history

Past tunnels (duration, owning token). Params: token (optional¹), protocol (filter, optional), limit (default 25).

list_regions

Available server regions. No auth required.

get_connection_info

Returns the CLI command (and, for load-balanced tunnels, a config file) to run the client manually — without spawning anything. Use in cloud sandboxes. Accepts the same arguments as create_tunnel.

Common Workflows

Expose a local web app

Copy code
create_tunnel(local_port=3000, protocol="http")
→ return public_url; close_tunnel(tunnel_id) when done

Custom subdomain

Copy code
create_tunnel(local_port=5173, protocol="http", subdomain="myapp-preview")

Database / TCP

Copy code
create_tunnel(local_port=5432, protocol="tcp")  → host:port to connect

UDP service

Copy code
create_tunnel(local_port=27015, protocol="udp")

Peer-to-peer — publish, then connect

Copy code
# Publisher (machine A)
create_tunnel(local_port=3000, protocol="p2p", secret="shared", peer_name="my-svc")
# Subscriber (machine B)
create_tunnel(local_port=8000, protocol="p2p", secret="shared", peer_target="my-svc")

Load-balanced pool with health checks

Run one create_tunnel per backend; members sharing (group, group_key) form one pool (requires the edge to have [load_balancing] enabled = true):

Copy code
create_tunnel(local_port=3000, protocol="http", subdomain="pool",
              group="web", group_key="shared-secret",
              health_check={ "type": "http", "path": "/health", "interval_secs": 10 })
create_tunnel(local_port=3001, protocol="http", subdomain="pool",
              group="web", group_key="shared-secret",
              health_check={ "type": "http", "path": "/health" })

Cloud sandbox (CLI fallback)

Copy code
get_connection_info(local_port=3000, protocol="http")
→ output the command for the user to run locally
→ list_tunnels() to confirm and fetch public_url

Prerequisites

The rustunnel CLI must be installed and on PATH (the MCP server spawns it):

bashCopy code
# Homebrew (macOS/Linux)
brew tap joaoh82/rustunnel && brew install rustunnel

# Or build from source
git clone https://github.com/joaoh82/rustunnel.git && cd rustunnel
make release   # builds rustunnel, rustunnel-mcp

Adding rustunnel to your harness

See docs/agent-integration.md for copy-paste MCP config for Claude Code, Claude Desktop, Codex, Cursor, Windsurf, Cline, and generic/custom agents — plus a one-command installer:

bashCopy code
curl -fsSL https://raw.githubusercontent.com/joaoh82/rustunnel/main/integrations/install.sh | bash

A minimal MCP config looks like:

jsonCopy code
{
  "mcpServers": {
    "rustunnel": {
      "command": "rustunnel-mcp",
      "args": ["--server", "eu.edge.rustunnel.com:4040", "--api", "https://eu.edge.rustunnel.com:8443"],
      "env": { "RUSTUNNEL_TOKEN": "<your-token>" }
    }
  }
}

Security Notes

• Tokens travel over HTTPS (use --insecure only in local dev with self-signed certs).
• MCP tools clean up subprocesses automatically; tunnels close when the MCP server exits.
• Protect the config file: chmod 600 ~/.rustunnel/config.yml.

Resources

• GitHub
• Agent Integration Guide
• MCP Server Docs
• Load Balancing
• P2P Tunnels