Beszel MCP Server

# Running Beszel MCP - All Methods Compared Quick comparison of different ways to run the Beszel MCP server. ## Method Comparison | Method | Speed | Setup | Best For | |--------|-------|-------|----------| | `uv run` | ⚡⚡⚡ | Instant | Development, Quick testing | | `uvx` | ⚡⚡⚡ | None | Claude Desktop, Production | | `pip + venv` | ⚡ | Manual | Traditional workflows | | `Docker` | ⚡⚡ | Complex | Containers, Cloud | ## 1. uv run (Recommended for Dev) **Fastest for development - no installation needed!** ```bash cd /path/to/beszel-mcp export BESZEL_URL="http://localhost:8090" uv run beszel-mcp ``` **Pros:** - ⚡ Instant - no `pip install` step - 🎯 No virtual env activation - 🔄 Automatic dependency management - 🚀 10-100x faster than pip **Cons:** - Requires uv installed **Best for:** Daily development, testing changes ## 2. uvx (Recommended for Claude Desktop) **Best for Claude Desktop integration** ```json { "mcpServers": { "beszel": { "command": "uvx", "args": ["--from", "/absolute/path/to/beszel-mcp", "beszel-mcp"], "env": { "BESZEL_URL": "http://localhost:8090" } } } } ``` **Pros:** - ⚡ Fast startup - 🔒 Isolated dependencies - 🎯 No global installation - 🔄 Auto-updates possible **Cons:** - Requires absolute path **Best for:** Claude Desktop, production use ## 3. Traditional pip + venv **Most compatible, works everywhere** ```bash cd /path/to/beszel-mcp python -m venv venv source venv/bin/activate pip install -e . beszel-mcp ``` **Pros:** - ✅ Works on all systems - 📦 Standard Python workflow - 🔧 Full control **Cons:** - 🐌 Slower installation (10-30 seconds vs 1-3) - 📝 Manual venv activation - 🔄 Manual dependency updates **Best for:** Systems without uv, corporate environments ## 4. Global Installation ```bash # With pip pip install /path/to/beszel-mcp # With uv uv tool install /path/to/beszel-mcp # Then run from anywhere beszel-mcp ``` **Pros:** - 🌍 Available globally - 📍 No path needed **Cons:** - ⚠️ Pollutes global Python - 🔄 Harder to update - 💥 Potential conflicts **Best for:** Personal single-user systems ## 5. Docker ```dockerfile FROM python:3.11-slim RUN curl -LsSf https://astral.sh/uv/install.sh | sh ENV PATH="/root/.cargo/bin:$PATH" WORKDIR /app COPY . . RUN uv sync CMD ["uv", "run", "beszel-mcp"] ``` **Pros:** - 📦 Fully isolated - ☁️ Cloud-ready - 🔒 Reproducible **Cons:** - 🐳 Requires Docker - 💾 Larger footprint - 🔧 More complex setup **Best for:** Containers, Kubernetes, cloud deployments ## Speed Comparison Installing FastMCP + httpx: ```bash # pip (traditional) time pip install fastmcp httpx # → 15-30 seconds # uv time uv pip install fastmcp httpx # → 1-3 seconds # uv run (no install needed!) time uv run beszel-mcp # → Instant after first run ``` ## Recommendation by Use Case ### Daily Development ```bash uv run beszel-mcp ``` **Why:** Instant, no setup, no venv activation ### Claude Desktop ```json "command": "uvx", "args": ["--from", "/path/to/beszel-mcp", "beszel-mcp"] ``` **Why:** Reliable, isolated, fast startup ### CI/CD Pipeline ```yaml - run: uv sync - run: uv run pytest ``` **Why:** Fast, reproducible, cacheable ### Production Server ```bash uv tool install /path/to/beszel-mcp # Or systemd with uv run ``` **Why:** Stable, managed, logged ### Corporate/Restricted Environment ```bash python -m venv venv source venv/bin/activate pip install -e . ``` **Why:** Standard, no special tools needed ## Configuration Files ### For uv (recommended) **pyproject.toml** already includes: ```toml [project.scripts] beszel-mcp = "beszel_mcp.__main__:main" [tool.uv] dev-dependencies = ["pytest>=8.0.0", "pytest-asyncio>=0.23.0"] ``` ### Environment Variables (all methods) ```bash export BESZEL_URL="http://localhost:8090" export BESZEL_EMAIL="admin@example.com" export BESZEL_PASSWORD="password" ``` Or create `.env`: ```env BESZEL_URL=http://localhost:8090 BESZEL_EMAIL=admin@example.com BESZEL_PASSWORD=password ``` ## Quick Decision Tree ``` Need it for Claude Desktop? ├─ Yes → Use uvx (Method 2) └─ No ├─ Developing/Testing? │ ├─ Have uv? → Use uv run (Method 1) │ └─ No uv? → Use pip+venv (Method 3) └─ Production? ├─ Docker/K8s? → Use Docker (Method 5) ├─ Systemd? → Use uv run with service └─ Simple? → Global install (Method 4) ``` ## Summary **For most users:** Start with `uv run` for development and `uvx` for Claude Desktop. **Installation time:** - uv: 10 seconds total (once) - pip: 30-60 seconds every time **Running time:** - uv run: Instant (after first run) - traditional: Instant (after venv activation) **Winner: uv** for speed, simplicity, and modern Python workflows! ⚡