mcp-server-catalog-reference
Navigate and recommend MCP servers from a curated catalog of 50 essential servers for AI agents with install commands and gotchas
How do I install this agent skill?
npx skills add https://github.com/reason-machines/mcp-skills --skill mcp-server-catalog-referenceIs this agent skill safe to install?
- Gen Agent Trust Hubpass
This skill acts as a curated reference catalog for Model Context Protocol (MCP) servers. It provides installation commands and configuration examples for various tools, along with comprehensive safety guidelines for handling sensitive operations such as database access, financial transactions, and blockchain interactions.
- Socketwarn
1 alert: gptAnomaly
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
MCP Server Catalog Reference
Skill by ara.so — MCP Skills collection.
What This Project Does
The 50-essential-mcp-servers catalog is a curated reference guide for MCP (Model Context Protocol) servers that work with Claude Code, Gemini CLI, and Codex. It provides:
- 50 vetted MCP servers organized into 9 categories (core dev, databases, ops/infra, productivity, fintech, web3, trading, creator tools, reasoning)
- Install commands for each agent (Claude Code, Codex, Gemini)
- Safety flags for servers that handle money, production data, or irreversible actions
- Practical selection advice to avoid tool bloat and performance degradation
The Three Critical Rules
Before recommending ANY MCP server, enforce these rules:
- Never recommend more than 5-7 servers — tool bloat makes agents slower and dumber
- Treat every server like untrusted code — pin versions, scope tokens read-only until proven
- Never enable writes to production — point at read replicas or snapshots for databases
Server Categories & Selection Guide
Core Development (Category 01)
Best for: Software engineers, full-stack developers Top picks: GitHub, Context7, Playwright, Filesystem, Brave Search
# GitHub (official, hosted)
claude mcp add github --transport http
# Context7 (official, hosted or local)
claude mcp add context7 --transport http
# OR local
claude mcp add context7 -- npx -y @context-labs/mcp-server-context7
# Playwright (official, local)
claude mcp add playwright -- npx -y @modelcontextprotocol/server-playwright
# Filesystem (official, local)
claude mcp add filesystem -- npx -y @modelcontextprotocol/server-filesystem
# Brave Search (official, local, requires API key)
claude mcp add brave-search -- npx -y @modelcontextprotocol/server-brave-search
Databases & Backend (Category 02)
Best for: Backend engineers, data teams Top picks: Postgres, Supabase, Neon, SQLite, Redis ⚠️ All flagged as "money or irreversible" — use read replicas
# Postgres (official archived, local)
claude mcp add postgres -- npx -y @modelcontextprotocol/server-postgres
# Config: Set DATABASE_URL env var to read replica
# Supabase (official, hosted)
claude mcp add supabase --transport http
# Requires project URL and anon key with restricted permissions
# Neon (official, hosted)
claude mcp add neon --transport http
# SQLite (community, local)
claude mcp add sqlite -- npx -y @benborla29/mcp-server-sqlite
# Redis (official, local)
claude mcp add redis -- npx -y @modelcontextprotocol/server-redis
Ops, Infra & Cloud (Category 03)
Best for: DevOps, SRE, platform engineers Top picks: Cloudflare, Vercel, AWS, Sentry ⚠️ Most flagged — can modify production infrastructure
# Cloudflare (official, hosted)
claude mcp add cloudflare --transport http
# Vercel (official, hosted)
claude mcp add vercel --transport http
# AWS (official, local)
claude mcp add aws -- npx -y @modelcontextprotocol/server-aws
# Requires AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY env vars
# Sentry (official, hosted, read-only)
claude mcp add sentry --transport http
# Kubernetes (community, local)
claude mcp add kubernetes -- npx -y mcp-server-kubernetes
# ⚠️ Can modify cluster state — scope RBAC carefully
Corporate Productivity (Category 04)
Best for: Product managers, team leads, knowledge workers Top picks: Slack, Linear, Notion, Jira, Google Drive
# Slack (community, local)
claude mcp add slack -- npx -y @modelcontextprotocol/server-slack
# ⚠️ Can send messages — review message drafts first
# Linear (official, hosted)
claude mcp add linear --transport http
# Notion (official, hosted)
claude mcp add notion --transport http
# Jira/Confluence (official, hosted)
claude mcp add jira --transport http
# Google Drive (official archived, local)
claude mcp add gdrive -- npx -y @modelcontextprotocol/server-gdrive
# Requires Google OAuth token with drive.readonly scope
Fintech & Payments (Category 05)
Best for: Financial applications, payment processing ⚠️ ALL servers in this category handle money — use sandbox/test mode
# Stripe (official, hosted or local)
claude mcp add stripe --transport http
# OR local with test key
claude mcp add stripe -- npx -y @modelcontextprotocol/server-stripe
# Set STRIPE_SECRET_KEY=sk_test_... (never live key)
# PayPal (official, hosted)
claude mcp add paypal --transport http
# Use sandbox credentials only
# Plaid (official, hosted)
claude mcp add plaid --transport http
# Use development environment
# QuickBooks (community, hosted)
# See vendor docs for OAuth setup
Web3 & Blockchain (Category 06)
Best for: Blockchain developers, dApp builders ⚠️ ALL servers execute irreversible on-chain transactions — use testnets
# Base (official, hosted)
claude mcp add base --transport http
# Configure for Base Sepolia testnet first
# Solana Agent Kit (community, local)
claude mcp add solana -- npx -y solana-agent-kit-mcp
# Set SOLANA_RPC_URL=https://api.devnet.solana.com
# Thirdweb (official, local)
claude mcp add thirdweb -- npx -y @thirdweb-dev/mcp-server
# Set THIRDWEB_SECRET_KEY and network to testnet
# The Graph (community, local, read-only)
claude mcp add the-graph -- npx -y @graphprotocol/mcp-server
# Dune (community, local, read-only)
claude mcp add dune -- npx -y dune-mcp-server
Trading & Markets (Category 07)
Best for: Traders, quant developers, market analysis ⚠️ Trading servers can place real orders — paper trade first
# Alpaca (official, local)
claude mcp add alpaca -- npx -y @alpacahq/mcp-server-alpaca
# Set ALPACA_API_KEY and APCA_API_BASE_URL=https://paper-api.alpaca.markets
# CCXT (community, local)
claude mcp add ccxt -- npx -y ccxt-mcp-server
# Configure exchange with testnet=true
# Polygon.io (official, local, read-only)
claude mcp add polygon -- npx -y @polygon-io/mcp-server
# CoinGecko (official, hosted, read-only)
claude mcp add coingecko --transport http
# TradingView (community, local, read-only)
claude mcp add tradingview -- npx -y tradingview-mcp
Creator Tools (Category 08)
Best for: Video editors, designers, content creators Top picks: Higgsfield, Figma, ElevenLabs, YouTube
# Higgsfield (official, hosted)
claude mcp add higgsfield --transport http
# DaVinci Resolve (community, local)
claude mcp add davinci -- npx -y davinci-resolve-mcp
# Figma (official, hosted or local)
claude mcp add figma --transport http
# ElevenLabs (official, local)
claude mcp add elevenlabs -- npx -y @elevenlabs/mcp-server
# Set ELEVENLABS_API_KEY env var
# YouTube (community, local)
claude mcp add youtube -- npx -y youtube-mcp-server
# Requires YouTube Data API key
Reasoning & Memory (Category 09)
Best for: Complex workflows, multi-step reasoning Official Anthropic servers for extended thinking
# Sequential Thinking (official, local)
claude mcp add sequential-thinking -- npx -y @modelcontextprotocol/server-sequential-thinking
# Memory (official, local)
claude mcp add memory -- npx -y @modelcontextprotocol/server-memory
Configuration Pattern
All three major agents use similar JSON config. Example for Claude Code's claude_desktop_config.json:
{
"mcpServers": {
"github": {
"transport": {
"type": "http",
"url": "https://github-mcp.glitch.me"
}
},
"postgres": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-postgres"],
"env": {
"DATABASE_URL": "postgresql://user:pass@read-replica.host:5432/db"
}
},
"stripe": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-stripe"],
"env": {
"STRIPE_SECRET_KEY": "${STRIPE_TEST_KEY}"
}
}
}
}
Codex Equivalent (~/.codex/config.toml)
[mcp.github]
transport = "http"
url = "https://github-mcp.glitch.me"
[mcp.postgres]
command = "npx"
args = ["-y", "@modelcontextprotocol/server-postgres"]
[mcp.postgres.env]
DATABASE_URL = "postgresql://user:pass@read-replica.host:5432/db"
Gemini CLI Equivalent (~/.gemini/settings.json)
{
"mcpServers": {
"github": {
"transport": "http",
"url": "https://github-mcp.glitch.me"
}
}
}
Selection Decision Tree
Use this logic when recommending servers:
1. What's the user's primary role?
- Software dev → Core dev (01) + Database (02)
- DevOps/SRE → Ops/infra (03) + Database (02)
- Product/PM → Productivity (04)
- Trader/quant → Trading (07) + Markets data
- Blockchain dev → Web3 (06) + Development (01)
- Content creator → Creator tools (08)
2. How many servers do they already have?
- 0-2: Recommend up to 3 more
- 3-5: Recommend 1-2 targeted additions
- 6+: STOP — suggest removing before adding
3. Does the server handle money/irreversible actions?
- YES: Warn explicitly, suggest sandbox/testnet, confirm human-in-loop
- NO: Proceed with install
4. Is there an official version?
- YES: Always prefer official over community
- NO: Check if community server is actively maintained
Safety Checklist for Dangerous Servers
Before installing servers flagged "money or irreversible":
Databases (Postgres, Supabase, Neon, Redis)
# ✅ GOOD: Read replica
DATABASE_URL="postgresql://readonly:pass@replica.host/db"
# ❌ BAD: Production primary with write access
DATABASE_URL="postgresql://admin:pass@prod-primary.host/db"
Fintech (Stripe, PayPal, Plaid)
# ✅ GOOD: Test mode
STRIPE_SECRET_KEY="sk_test_..."
PAYPAL_MODE="sandbox"
# ❌ BAD: Live keys
STRIPE_SECRET_KEY="sk_live_..."
Web3 (Base, Thirdweb, Solana)
# ✅ GOOD: Testnet
SOLANA_RPC_URL="https://api.devnet.solana.com"
BASE_NETWORK="base-sepolia"
# ❌ BAD: Mainnet without confirmation
SOLANA_RPC_URL="https://api.mainnet-beta.solana.com"
Trading (Alpaca, CCXT)
# ✅ GOOD: Paper trading
APCA_API_BASE_URL="https://paper-api.alpaca.markets"
CCXT_EXCHANGE_SANDBOX="true"
# ❌ BAD: Live trading without testing
APCA_API_BASE_URL="https://api.alpaca.markets"
Common Patterns
Pattern 1: Dev Stack (5 servers)
claude mcp add github --transport http
claude mcp add context7 --transport http
claude mcp add postgres -- npx -y @modelcontextprotocol/server-postgres
claude mcp add sentry --transport http
claude mcp add playwright -- npx -y @modelcontextprotocol/server-playwright
Pattern 2: Content Creator Stack (5 servers)
claude mcp add higgsfield --transport http
claude mcp add figma --transport http
claude mcp add elevenlabs -- npx -y @elevenlabs/mcp-server
claude mcp add youtube -- npx -y youtube-mcp-server
claude mcp add davinci -- npx -y davinci-resolve-mcp
Pattern 3: Web3 Builder Stack (5 servers, testnet-safe)
claude mcp add github --transport http
claude mcp add base --transport http # Configure for Sepolia
claude mcp add the-graph -- npx -y @graphprotocol/mcp-server
claude mcp add dune -- npx -y dune-mcp-server
claude mcp add context7 --transport http
Pattern 4: Product/PM Stack (5 servers)
claude mcp add linear --transport http
claude mcp add notion --transport http
claude mcp add slack -- npx -y @modelcontextprotocol/server-slack
claude mcp add gdrive -- npx -y @modelcontextprotocol/server-gdrive
claude mcp add github --transport http
Troubleshooting
Server Not Appearing in Tool List
# 1. Verify config syntax
cat ~/.config/Claude/claude_desktop_config.json | jq .
# 2. Check server can start independently
npx -y @modelcontextprotocol/server-postgres
# 3. Restart agent completely
# Claude Code: Cmd+Q and reopen
# Codex: codex restart
Agent Getting Slow/Confused
# Count your servers
jq '.mcpServers | keys | length' ~/.config/Claude/claude_desktop_config.json
# If > 7: Remove least-used servers
# Keep only what you used this week
Permission/Auth Errors
# For hosted servers: Re-authenticate
claude mcp remove github
claude mcp add github --transport http # Triggers OAuth
# For local servers: Check env vars are set
npx -y @modelcontextprotocol/server-stripe # Will fail if STRIPE_SECRET_KEY missing
Server Version Conflicts
# Pin to specific version
claude mcp add postgres -- npx -y @modelcontextprotocol/server-postgres@1.2.3
# Check installed version
npm list -g @modelcontextprotocol/server-postgres
Dangerous Action Attempted
# If agent tries to execute money/irreversible action:
# 1. STOP immediately
# 2. Review the exact command/API call
# 3. Verify environment is test/sandbox
# 4. If prod: Remove server or switch to read-only
# Example: Switch Stripe to test
claude mcp remove stripe
# Edit config to use sk_test_ key
claude mcp add stripe -- npx -y @modelcontextprotocol/server-stripe
Quick Reference: Server Safety Matrix
| Server | Read-Only? | Requires Sandbox? | Production Risk |
|---|---|---|---|
| GitHub | ✅ Can be | ❌ No | Low (code reviews exist) |
| Postgres | ❌ No | ✅ Yes | HIGH (data loss) |
| Stripe | ❌ No | ✅ Yes | CRITICAL (money) |
| Slack | ❌ No | ❌ No | Medium (can spam) |
| Base/Solana | ❌ No | ✅ Yes | CRITICAL (irreversible) |
| Alpaca | ❌ No | ✅ Yes | CRITICAL (real trades) |
| Sentry | ✅ Yes | ❌ No | None |
| CoinGecko | ✅ Yes | ❌ No | None |
Environment Variables Reference
# Database servers
DATABASE_URL="postgresql://user:pass@host:5432/db"
REDIS_URL="redis://localhost:6379"
# Fintech
STRIPE_SECRET_KEY="sk_test_..."
PLAID_CLIENT_ID="..."
PLAID_SECRET="..."
PLAID_ENV="sandbox"
# Web3
SOLANA_RPC_URL="https://api.devnet.solana.com"
SOLANA_PRIVATE_KEY="..."
THIRDWEB_SECRET_KEY="..."
THIRDWEB_NETWORK="base-sepolia"
# Trading
ALPACA_API_KEY="..."
ALPACA_SECRET_KEY="..."
APCA_API_BASE_URL="https://paper-api.alpaca.markets"
# APIs
BRAVE_SEARCH_API_KEY="..."
ELEVENLABS_API_KEY="..."
YOUTUBE_API_KEY="..."
COINGECKO_API_KEY="..."
# Cloud providers
AWS_ACCESS_KEY_ID="..."
AWS_SECRET_ACCESS_KEY="..."
AWS_REGION="us-east-1"
CLOUDFLARE_API_TOKEN="..."
Key Recommendations for AI Agents
When a user asks for MCP server help:
- Always ask their role first — don't guess
- Count their existing servers — enforce the 5-7 limit
- Warn on every dangerous server — be explicit about risks
- Provide the exact install command — don't make them hunt
- Suggest test/sandbox by default — never assume prod is okay
- Recommend removal — if adding server #8, suggest dropping #1
- Link to official docs — catalog is a map, not the territory
The catalog exists because the MCP ecosystem is huge and dangerous. Your job is to be the safety rail between "sounds cool" and "lost money/data."
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/reason-machines/mcp-skills/mcp-server-catalog-reference">View mcp-server-catalog-reference on skillZs</a>