Essential MCP Servers & Tooling for Claude Code CLI

MCP servers are the new standard for extending Claude Code CLI, offering robust tooling for development, automation, and cloud operations, which can be configured on a self-hosted Ubuntu VPS for enhanced control and security.

| 3 min read
MCP servers are the new standard for extending Claude Code CLI, offering robust tooling for development, automation, and cloud operations, which can be configured on a self-hosted Ubuntu VPS for enhanced control and security.

Since you're on a fresh Ubuntu VPS with Claude Code CLI, the modern extension paradigm has shifted from traditional "plugins" to MCP (Model Context Protocol) servers. Claude Code natively supports MCP, so you'll extend it by running local MCP server processes and pointing Claude to them via config.

🔹 Official Anthropic MCP servers: Check if already have. (Claude Code CLI auto-bundles or auto-detects the official suite like filesystem, git, fetch, etc. Verify with claude mcp list or claude config.)

Below are the most reliable, actively maintained, and trending third-party/community MCP servers & tooling for development, automation, and agentic workflows as of 2025–2026.

🧠 1. Development & Code Workflow

MCP Server Purpose Install/Run
mcp-server-github (community) PRs, issues, repo navigation, code search npx -y @anthropic/mcp-server-github (or community forks like @smol-ai/mcp-server-github)
mcp-server-docker Build, run, inspect containers & logs npx -y @modelcontextprotocol/server-docker
mcp-server-postgres / mcp-server-sqlite Direct DB querying, schema inspection npx -y @modelcontextprotocol/server-postgres
mcp-server-playwright Headless browser automation for testing/scraping npx -y @anthropic/mcp-server-playwright or @playwright/mcp
aider (standalone CLI) Git-aware AI coding assistant with shell integration pipx install aider-chat (pairs well alongside Claude Code)

🤖 2. Agentic & Automation

MCP Server Purpose Install/Run
mcp-server-terminal / mcp-server-shell Safe, scoped command execution with context npx -y @anthropic/mcp-server-terminal or community mcp-server-shell
firecrawl-mcp Web scraping, content extraction, JS rendering npx -y @mendable/firecrawl-mcp
mcp-server-linear / mcp-server-jira / mcp-server-notion Task tracking, sprint automation, doc sync npx -y @anthropic/mcp-server-linear (or @linear/mcp)
mcp-server-slack / mcp-server-discord Notification routing, bot-style workflows npx -y @anthropic/mcp-server-slack
langgraph-mcp or crewai-mcp Bridge multi-agent frameworks into Claude CLI pip install langgraph-mcp or crewai[tools]

☁️ 3. Infrastructure & Cloud Ops

MCP Server Purpose Install/Run
mcp-server-aws / mcp-server-gcp Cloud resource management, logs, IAM, S3/GCS npx -y @anthropic/mcp-server-aws
mcp-server-kubernetes Cluster ops, pod logs, deployments, HPA npx -y @anthropic/mcp-server-kubernetes
mcp-server-sentry Error tracking, release monitoring npx -y @anthropic/mcp-server-sentry
mcp-server-supabase / mcp-server-vercel Platform-specific devops & deployments npx -y @anthropic/mcp-server-supabase

🛠️ 4. MCP Management & Discovery (Trending 2025–2026)

Tool Purpose Install
smithery-cli npm-style package manager for MCP servers npm i -g @smithery/cli
glama-cli MCP registry & version manager npm i -g @glama/cli
mcp-proxy Route, auth, and sandbox MCP connections go install github.com/modelcontextprotocol/mcp-proxy@latest
mcp-inspector Debug & visualize MCP server interactions npx -y @modelcontextprotocol/inspector

📦 Ubuntu VPS Setup & Configuration

  1. Install runtimes (MCP servers use Node, Python, or Go):

    sudo apt update
sudo apt install -y nodejs npm python3-pip python3-venv git curl jq
curl -fsSL https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bash
nvm install 20 && nvm use 20
pip3 install pipx && pipx ensurepath

  2. Configure Claude Code CLI to use MCP: Create or edit ~/.claude/settings.json (or ~/.config/claude-code/mcp.json depending on your version):

    {
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@anthropic/mcp-server-github"],
      "env": { "GITHUB_TOKEN": "your_token_here" }
    },
    "firecrawl": {
      "command": "npx",
      "args": ["-y", "@mendable/firecrawl-mcp"],
      "env": { "FIRECRAWL_API_KEY": "your_key_here" }
    },
    "docker": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-docker"]
    }
  }
}

    Restart Claude Code CLI or run claude mcp reload.

🔒 VPS & Security Best Practices

  • Run as non-root: Create a dedicated user (adduser claude-dev) and run MCP servers under it.
  • Scope permissions: Use env vars for tokens, never hardcode. Restrict filesystem MCP to specific dirs.
  • Sandbox execution: Prefer mcp-proxy or run MCP servers in systemd user services with PrivateTmp=true, RestrictSUIDSGID=true.
  • Resource limits: MCP servers are lightweight, but headless browsers (Playwright) need RAM. Add swap if needed: fallocate -l 2G /swapfile && mkswap /swapfile && swapon /swapfile.

📈 Trend Context (2025–2026)

  • MCP is the de facto extension standard for CLI AI tools. Traditional IDE plugins are deprecated in favor of stateless, local MCP servers.
  • Agentic CLI workflows now rely on MCP + safe shell wrappers + local execution engines. Tools like smithery and glama act as the "npm/pip" for the MCP ecosystem.
  • Self-hosted VPS advantage: You can run MCP servers persistently, secure them with ufw/fail2ban, and scale them via systemd or Docker Compose without rate limits or cloud vendor lock-in.

For source verification & active repos, check:

  • GitHub: github.com/modelcontextprotocol/servers
  • MCP Registry: smithery.ai & glama.ai
  • Anthropic MCP Docs: docs.anthropic.com/en/docs/mcp

Let me know your primary stack (web, data, infra, etc.) and I'll tailor a minimal, production-ready MCP config + systemd templates for your VPS.