The MCP Ecosystem — Essential Servers, Setup Guides & Cross-Tool Patterns

The complete guide to Model Context Protocol (MCP) — the universal standard that connects AI coding agents to your tools, databases, APIs, and services. This cookbook covers the most popular MCP servers, step-by-step setup for Claude Code, Cursor, and Codex, and the agentic coding patterns that make MCP indispensable.

MCP is the "USB-C of AI integrations" — an open protocol by Anthropic that standardizes how AI models communicate with external systems. Instead of custom glue code for every tool, MCP provides one universal protocol. With 28,000+ servers indexed and adoption by OpenAI, Google, Microsoft, and every major AI coding tool, MCP is now the standard for agentic development.

---

What is MCP

Model Context Protocol (MCP) is an open standard created by Anthropic in November 2024. It provides a universal way for AI applications to connect with external tools, data sources, and services through a standardized JSON-RPC 2.0 protocol. Why MCP Matters:

• Dynamic tool discovery at runtime
• Standardized invocation across all tools
• Context persistence across conversations
• Composability — run multiple servers simultaneously Architecture:

MCP Host (Claude Code, Cursor, Codex)
  ├── MCP Client (Server A) → MCP Server (GitHub)
  └── MCP Client (Server B) → MCP Server (Supabase)

Two Transport Types:

Transport	How It Works	Best For
STDIO	Host spawns server as child process, communicates via stdin/stdout	Local tools (filesystem, databases)
Streamable HTTP	Server exposes HTTP endpoint, supports OAuth	Remote/hosted services (Sentry, Linear, Notion)
Three Primitives:
Primitive	Description	Example
---	---	---
Tools	Executable functions AI can call	create_issue, query_database
Resources	Data AI can read	Database schemas, file contents
Prompts	Reusable prompt templates	System prompts, few-shot examples

---

The 15 Most Popular MCP Servers

1. Filesystem MCP Server

Controlled read/write access to your local filesystem. Claude Code:

claude mcp add filesystem -s user -- npx -y @modelcontextprotocol/server-filesystem ~/Projects

Cursor / Claude Desktop (JSON):

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/dir"]
    }
  }
}

Codex (TOML):

[mcpServers.filesystem]
command = "npx"
args = ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/dir"]

Tools: read_file, write_file, list_directory, search_files, move_file

---

2. GitHub MCP Server (Official)

Full GitHub API — repos, PRs, issues, code search, file operations. Claude Code:

claude mcp add github -- npx -y @modelcontextprotocol/server-github
export GITHUB_PERSONAL_ACCESS_TOKEN=ghp_your_token

Cursor (JSON):

{
  "mcpServers": {
    "github": {
      "command": "docker",
      "args": ["run", "-i", "--rm", "-e", "GITHUB_PERSONAL_ACCESS_TOKEN", "ghcr.io/github/github-mcp-server"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_token" }
    }
  }
}

Codex:

codex mcp add github -- npx -y @modelcontextprotocol/server-github

Key tools: create_repository, create_pull_request, search_code, list_commits, merge_pull_request

---

3. PostgreSQL MCP Server

Database queries, schema inspection, and performance analysis. Config (all tools):

{
  "mcpServers": {
    "postgres": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres", "postgresql://user:pass@localhost:5432/db"]
    }
  }
}

Security: Always use read-only database credentials in production.

---

4. Supabase MCP Server

Comprehensive Supabase management — database, migrations, RLS, Edge Functions. Claude Code:

claude mcp add supabase -e SUPABASE_ACCESS_TOKEN=token -- npx -y @supabase/mcp-server-supabase@latest --project-ref ref

Cursor (JSON):

{
  "mcpServers": {
    "supabase": {
      "command": "npx",
      "args": ["-y", "@supabase/mcp-server-supabase@latest", "--project-ref", "ref"],
      "env": { "SUPABASE_ACCESS_TOKEN": "token" }
    }
  }
}

Tip: Always use --read-only for production. Never connect to production data.

---

5. Playwright MCP Server (Microsoft)

Browser automation — navigate, click, fill forms, screenshot, monitor network. Config (all tools):

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["@playwright/mcp@latest"]
    }
  }
}

Use case: Visual UI feedback — agent opens your app, sees the UI, and self-corrects.

---

6. Figma MCP Server

Bridge between design files and code — extract specs, tokens, and component structures. Claude Code:

claude mcp add figma -s user -- npx -y @figma/mcp-server@latest

Cursor:

{
  "mcpServers": {
    "figmaRemoteMcp": {
      "url": "https://mcp.figma.com/mcp"
    }
  }
}

---

7. Brave Search MCP Server

Web search from within coding sessions (2,000 free queries/month). Config:

{
  "mcpServers": {
    "brave-search": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-brave-search"],
      "env": { "BRAVE_API_KEY": "key" }
    }
  }
}

---

8. Memory MCP Server

Persistent knowledge graph — remembers entities, preferences, and decisions across sessions. Config:

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-memory"]
    }
  }
}

Tools: create_entities, create_relations, search_nodes, read_graph

---

9. Sequential Thinking MCP Server

The number 1 most-used MCP server (5,550+ monthly uses on Smithery). Forces step-by-step reasoning with branching and revision support. Config:

{
  "mcpServers": {
    "sequential-thinking": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]
    }
  }
}

Best for: Complex debugging, architecture decisions, multi-file refactoring.

---

10. Context7 MCP Server

Real-time, up-to-date documentation for libraries and frameworks (46k GitHub stars). Claude Code:

claude mcp add context7 -- npx -y @upstash/context7-mcp@latest

Codex:

codex mcp add context7 -- npx -y @upstash/context7-mcp

---

11. Slack MCP Server

Post messages, read channels, list users, add reactions.

12. Sentry MCP Server

Error analysis, stack traces, releases, root cause detection.

claude mcp add --transport http sentry https://mcp.sentry.dev/mcp

13. Linear MCP Server

Project tracking, bug triage, release management.

# Codex
url = "https://mcp.linear.app/mcp"

14. Docker MCP Server

Container lifecycle management from within AI sessions.

15. Notion MCP Server

Read/write Notion pages, databases, and project documentation.

claude mcp add --transport http notion https://mcp.notion.com/mcp

---

MCP Registries — Where to Find Servers

Registry	URL	Notes
Official MCP Registry	registry.modelcontextprotocol.io	Canonical source
Smithery	smithery.ai	One-click installs, usage stats
Glama	glama.ai/mcp/servers	Curated, security-audited
MCP.so	mcp.so	Search engine for servers
PulseMCP	pulsemcp.com	News and discovery
Cursor Directory	cursor.directory	Cursor-specific configurations
awesome-mcp-servers	github.com/punkpeye/awesome-mcp-servers	81k stars, community curated

---

Building Custom MCP Servers

Python (FastMCP)

from fastmcp import FastMCP

mcp = FastMCP("my-custom-server")

@mcp.tool()
def get_weather(city: str) -> str:
    """Get current weather for a city."""
    return f"Weather in {city}: 72°F, sunny"

if __name__ == "__main__":
    mcp.run(transport="stdio")

TypeScript

import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';

const server = new Server({ name: 'my-server', version: '1.0.0' });
// Register tools...
const transport = new StdioServerTransport(server);
transport.listen();

---

Agentic Coding Patterns

Pattern 1: Context Files (CLAUDE.md / AGENTS.md)

Every project should have a context file that gives AI agents persistent project knowledge — architecture, conventions, commands, and gotchas.

Tool	File	Discovery
Claude Code	CLAUDE.md	Auto-loaded at session start
Codex CLI	AGENTS.md	Read before every task
Cursor	.cursor/rules/*.mdc	Auto-attached by file pattern

Pattern 2: Test-Driven AI Development

Write failing tests first → let the AI implement until tests pass. This is the single most reliable pattern for AI-generated code.

Pattern 3: Plan → Review → Execute

Always have the AI plan before coding. Review the plan, push back, then approve.

Pattern 4: Multi-Agent Fan-Out

Split large tasks into independent subtasks and run multiple agents in parallel.

Pattern 5: PR Review Automation

Integrate AI review into CI/CD — every PR gets automated analysis for bugs, security, and conventions.

Pattern 6: Context Management

• Start new sessions for unrelated tasks
• Use /compact when context gets heavy
• Reference specific files instead of entire codebases
• Break large prompts into focused steps

---

Cross-Tool Configuration Quick Reference

Claude Code

claude mcp add <name> -- <command>
claude mcp add --transport http <name> <url>
claude mcp list

Cursor

// .cursor/mcp.json or ~/.cursor/mcp.json
{
  "mcpServers": {
    "name": { "command": "npx", "args": ["..."] }
  }
}

Codex CLI

# ~/.codex/config.toml
[mcp_servers.name]
command = "npx"
args = ["-y", "@package/server"]

---

Recommended Starter Stack

For developers setting up their first MCP-powered agentic workflow:

Server	Why It's Essential
Context7	Eliminates hallucinated API docs
GitHub	Full PR and issue workflow from agent
Playwright	Visual verification loop
Supabase or PostgreSQL	Database-aware code generation
Memory	Persistent context across sessions
Sequential Thinking	Better reasoning for complex tasks

---

Last updated: February 27, 2026 Built for authorityaitools.com — AI Coding Tools Directory