Best MCP Servers for Claude Code: 18 Tools You Need in 2026

The MCP ecosystem has exploded. Since Anthropic open-sourced the Model Context Protocol in late 2024, the community has built over 10,000 public MCP servers, and the official SDKs have surpassed 97 million downloads across TypeScript and Python. Every major AI tool vendor now supports MCP — Anthropic, OpenAI, Google, and Microsoft have all adopted it as the standard for connecting AI agents to external tools.

But with thousands of servers available, finding the ones that actually matter is the hard part.

This guide cuts through the noise. Here are the 18 MCP servers that deliver the most value for Claude Code users in 2026, organized by category, with install commands you can copy-paste right now.

Quick Start: How to Install MCP Servers in Claude Code

Before diving into the list, here is how MCP server installation works in Claude Code.

The Basic Command

claude mcp add <name> <command> [args...]

Adding with Environment Variables

claude mcp add <name> -e KEY=value <command> [args...]

Scope Options

# Available only in the current project (default)
claude mcp add <name> npx @package/server

# Available across all your projects
claude mcp add --scope user <name> npx @package/server

Managing Servers

# List all configured servers
claude mcp list

# Remove a server
claude mcp remove <name>

# Check server status inside Claude Code
/mcp

Rule of thumb: Use --scope user for general-purpose tools you want everywhere (GitHub, Context7, browser). Use project scope for project-specific connections (databases, internal APIs).

For a deeper walkthrough of MCP setup, configuration, and building your own servers, see our Claude Code MCP Setup Guide.

Database Servers

1. PostgreSQL

What it does: Connects Claude Code directly to your PostgreSQL database. Claude can inspect schemas, write SQL queries, and return results — all without you switching to a database client.

Install:

claude mcp add postgres \
  npx @anthropic/mcp-postgres \
  "postgresql://user:pass@localhost:5432/mydb"

Key features:

• Read-only query execution by default (safe for production databases)
• Automatic schema introspection — Claude understands your tables and columns
• Support for parameterized queries
• Connection string passed as argument, not stored in plain text config

When to use it: Debugging data issues, generating reports, exploring unfamiliar databases, verifying migrations. Pair it with a read-only database user for production safety.

2. SQLite

What it does: Provides SQLite database management — querying, schema inspection, and data manipulation for local SQLite files.

Install:

claude mcp add sqlite \
  npx @anthropic/mcp-sqlite \
  /path/to/database.db

Key features:

• Full read/write access to SQLite databases
• Schema exploration and table management
• Works with any local .db file
• Lightweight — no database server required

When to use it: Working with mobile app databases, local development databases, embedded systems data, or any project using SQLite as its storage layer.

3. MySQL

What it does: Connects Claude Code to MySQL and MariaDB databases with the same query and schema inspection capabilities as the PostgreSQL server.

Install:

claude mcp add mysql \
  npx @anthropic/mcp-mysql \
  "mysql://user:pass@localhost:3306/mydb"

Key features:

• Schema introspection for MySQL-specific types
• Support for stored procedures and views
• Read-only mode available for production safety

When to use it: Any project backed by MySQL or MariaDB. Essential for legacy application maintenance where MySQL is the primary data store.

Browser Automation

4. Playwright (Microsoft Official)

What it does: Gives Claude Code full browser automation capabilities through Microsoft’s Playwright engine. Claude can navigate pages, click elements, fill forms, take screenshots, and run end-to-end test flows.

Install:

claude mcp add playwright \
  npx @anthropic/mcp-playwright

Key features:

• Cross-browser support: Chromium, Firefox, WebKit
• Uses accessibility tree snapshots for element interaction (more reliable than CSS selectors)
• Screenshot capture for visual verification
• Network request interception and monitoring
• Runs entirely on your local machine — no data leaves your environment

When to use it: Writing and debugging E2E tests, scraping web data, automating repetitive browser workflows, verifying UI changes after code modifications. This is the go-to choice for serious browser automation.

For a deeper look at browser automation options, see our browser automation guide.

5. Fetch

What it does: A simpler alternative to Playwright for basic web content retrieval. Fetches URLs and converts HTML to markdown that Claude can process.

Install:

claude mcp add fetch \
  npx @anthropic/mcp-fetch

Key features:

• HTML to markdown conversion for clean text extraction
• Handles redirects and authentication headers
• Lightweight — no browser engine required
• Supports custom headers for authenticated endpoints

When to use it: When you need to read web content without full browser interaction. API documentation lookups, checking deployment status pages, or pulling content from URLs shared in conversations.

Code & Version Control

6. GitHub (Official by GitHub)

What it does: Full GitHub integration — issues, pull requests, repository search, branch management, CI/CD workflows, and code review. This is GitHub’s own official MCP server, not a third-party wrapper.

Install:

claude mcp add github \
  -e GITHUB_PERSONAL_ACCESS_TOKEN=ghp_your_token \
  npx @anthropic/mcp-github

Key features:

• Create, read, and update issues and pull requests
• Search across repositories and code
• Trigger and monitor GitHub Actions workflows
• Branch and tag management
• Review and merge pull requests with comments

When to use it: Any project hosted on GitHub. This eliminates context-switching to the browser for issue tracking, PR reviews, and CI monitoring. Use --scope user since you likely want this everywhere.

7. Git

What it does: Provides Git operations beyond what Claude Code’s built-in shell access offers — structured access to repository history, diffs, blame, and branch operations with proper error handling.

Install:

claude mcp add git \
  npx @anthropic/mcp-git \
  /path/to/repository

Key features:

• Structured log and diff output (easier for Claude to parse than raw git output)
• Blame analysis for tracking down code authorship
• Branch comparison and merge-base detection
• Commit history search by message, author, or date range

When to use it: When you need Claude to do deep repository archaeology — understanding code evolution, finding when bugs were introduced, or analyzing contribution patterns across a large codebase.

Documentation & Knowledge

8. Context7

What it does: Fetches up-to-date library documentation at query time. When Claude needs to use an API it is unsure about, Context7 pulls the latest official docs instead of relying on training data that may be months or years old.

Install:

claude mcp add context7 \
  --scope user \
  npx @upstash/context7-mcp

Key features:

• Version-specific documentation retrieval
• Covers thousands of popular libraries and frameworks
• Two tools: resolve-library-id (finds the correct library) and query-docs (fetches documentation)
• Eliminates hallucinated APIs — Claude works from real documentation

When to use it: Always. This should be one of your default --scope user servers. It prevents Claude from inventing API calls that do not exist, especially for fast-moving libraries like Next.js, React, or LangChain where APIs change frequently.

Pro tip: When asking Claude about a library, add “use context7” to your prompt. Claude will automatically look up the latest docs before writing code.

9. Memory (Knowledge Graph)

What it does: Gives Claude persistent memory across conversations using a local knowledge graph. Claude can store and retrieve entities, relationships, and observations — remembering context about your project, preferences, and decisions.

Install:

claude mcp add memory \
  npx @anthropic/mcp-memory

Key features:

• Local knowledge graph stored as JSON
• Entities with typed relationships (person, project, concept)
• Observations attached to entities
• Persists across Claude Code sessions
• Fully local — no data sent to external services

When to use it: Long-running projects where you want Claude to remember architectural decisions, team member roles, coding conventions, or domain-specific terminology without re-explaining every session.

AI Reasoning

10. Sequential Thinking

What it does: Introduces a structured, multi-step reasoning process for Claude. Instead of jumping to a solution, Claude breaks problems into sequential thought steps with the ability to branch, revise, and verify along the way.

Install:

claude mcp add sequential-thinking \
  npx @modelcontextprotocol/server-sequential-thinking

Key features:

• Step-by-step thought decomposition
• Ability to revise previous reasoning steps
• Branching for exploring alternative approaches
• Configurable depth — set minimum and maximum thought steps
• Thought logging can be disabled for cleaner output (set DISABLE_THOUGHT_LOGGING=true)

When to use it: Complex architectural decisions, debugging multi-system issues, planning large refactors, or any problem where the first intuition might be wrong. This server makes Claude’s reasoning visible and auditable.

Productivity & Communication

11. Notion

What it does: Connects Claude Code to your Notion workspace. Claude can search pages, read content, create new pages, update existing ones, and manage databases.

Install:

claude mcp add notion \
  -e NOTION_API_KEY=ntn_your_key \
  npx @anthropic/mcp-notion

Key features:

• Search across your entire Notion workspace
• Read and write page content (including rich text blocks)
• Create pages in specific databases
• Query database entries with filters and sorts
• Comment on pages for team collaboration

When to use it: Teams that use Notion for documentation, project management, or knowledge bases. Claude can pull requirements from Notion, update status pages after deployments, or create documentation pages from code analysis.

12. Slack

What it does: Lets Claude Code interact with your Slack workspace — read messages, post updates, search conversations, and manage channels.

Install:

claude mcp add slack \
  -e SLACK_BOT_TOKEN=xoxb-your-token \
  -e SLACK_TEAM_ID=T01234567 \
  npx @anthropic/mcp-slack

Key features:

• Read and send messages in channels and DMs
• Search message history across your workspace
• Post formatted messages with attachments
• Channel management (list, join, create)
• Thread support for organized discussions

When to use it: Automated notifications when Claude completes long tasks, pulling context from team discussions, or posting deployment summaries. Pairs well with Claude Code Hooks for fully automated workflows.

13. Linear

What it does: Integrates with Linear for issue tracking and project management. Claude can create issues, update statuses, search for existing tickets, and manage project workflows.

Install:

claude mcp add linear \
  -e LINEAR_API_KEY=lin_api_your_key \
  npx @anthropic/mcp-linear

Key features:

• Create and update issues with labels, priorities, and assignees
• Search issues by text, status, or assignee
• Project and cycle management
• Team and member lookups
• Workflow state transitions

When to use it: Engineering teams using Linear for project management. Claude can automatically create bug tickets from code analysis, update issue status after completing work, or search for related tickets before starting a new feature.

File System & Infrastructure

14. Filesystem

What it does: The official MCP server for secure local file operations. While Claude Code has built-in file access, this server adds granular permission controls, directory tree visualization, and advanced search capabilities.

Install:

claude mcp add filesystem \
  npx @anthropic/mcp-filesystem \
  /path/to/allowed/directory

Key features:

• Read, write, and edit files within allowed directories
• Directory tree visualization
• File search by name or content patterns
• Granular permission boundaries — only the directories you specify
• Symlink-aware operations

When to use it: When you need strict filesystem sandboxing or when working across multiple project directories. The directory argument acts as a security boundary — Claude cannot access files outside the specified paths.

15. Docker

What it does: Manages Docker containers, images, volumes, and networks directly from Claude Code. Claude can build images, start containers, inspect logs, and manage Docker Compose stacks.

Install:

claude mcp add docker \
  npx @anthropic/mcp-docker

Key features:

• Container lifecycle management (create, start, stop, remove)
• Image building and management
• Log inspection and container exec
• Docker Compose stack operations
• Volume and network management

When to use it: Microservice development, debugging container issues, managing local development environments, or automating Docker workflows. Especially useful when you need to inspect running containers while debugging application issues.

Cloud & Infrastructure

16. AWS Documentation

What it does: Provides Claude Code with access to official AWS documentation. Claude can search and read AWS service docs, getting accurate configuration details, API references, and best practices.

Install:

claude mcp add aws-docs \
  npx @awslabs/aws-documentation-mcp-server

Key features:

• Search across all AWS service documentation
• Read specific documentation pages
• Service-specific recommendations
• Always up-to-date — pulls from live AWS docs

When to use it: Any project using AWS services. Eliminates guesswork about IAM policies, CloudFormation templates, Lambda configurations, and other AWS-specific details where getting the syntax exactly right matters.

17. Cloudflare

What it does: Manages Cloudflare Workers, KV stores, R2 buckets, and DNS records. Claude can deploy edge functions, configure caching rules, and manage your Cloudflare infrastructure.

Install:

claude mcp add cloudflare \
  -e CLOUDFLARE_API_TOKEN=your_token \
  npx @anthropic/mcp-cloudflare

Key features:

• Workers deployment and management
• KV namespace operations (read, write, list, delete)
• R2 object storage management
• DNS record configuration
• Zone and domain management

When to use it: Projects deployed on Cloudflare’s edge platform. Claude can deploy Workers, update KV stores, manage R2 buckets, and configure DNS — all without leaving the terminal.

Monitoring & Observability

18. Sentry

What it does: Connects Claude Code to your Sentry error tracking. Claude can search for errors, inspect stack traces, analyze error trends, and help debug production issues with full context.

Install:

claude mcp add sentry \
  -e SENTRY_AUTH_TOKEN=your_token \
  -e SENTRY_ORG=your-org \
  npx @anthropic/mcp-sentry

Key features:

• Search and filter error events
• Inspect detailed stack traces with source context
• Analyze error frequency and trends
• Link errors to specific releases and commits
• Issue assignment and status management

When to use it: Debugging production errors. Instead of switching to the Sentry dashboard, Claude can pull error details, analyze the stack trace alongside your codebase, and suggest fixes with full context of both the error and the code.

My Recommended Setup

Here is the MCP configuration I use daily across all projects. These five servers cover 90% of typical development needs:

# Universal tools (user scope — available everywhere)
claude mcp add --scope user github \
  -e GITHUB_PERSONAL_ACCESS_TOKEN=ghp_xxx \
  npx @anthropic/mcp-github

claude mcp add --scope user context7 \
  npx @upstash/context7-mcp

claude mcp add --scope user playwright \
  npx @anthropic/mcp-playwright

claude mcp add --scope user sequential-thinking \
  npx @modelcontextprotocol/server-sequential-thinking

claude mcp add --scope user fetch \
  npx @anthropic/mcp-fetch

Then per-project, I add whatever the project needs:

# Project-specific database
claude mcp add postgres \
  npx @anthropic/mcp-postgres \
  "postgresql://dev:password@localhost:5432/myapp"

# Project-specific integrations
claude mcp add slack \
  -e SLACK_BOT_TOKEN=xoxb-xxx \
  -e SLACK_TEAM_ID=T01234567 \
  npx @anthropic/mcp-slack

Verify everything is connected:

claude mcp list

Inside Claude Code, run /mcp to confirm all servers are running and their tools are available.

How to Choose the Right MCP Servers

Not every project needs all 18 servers. Here is a decision framework:

Start small. Install two or three servers that match your daily workflow, learn how they work, and add more as you identify gaps. Running too many servers that you rarely use just adds startup overhead.

Where to Find More MCP Servers

The 18 servers in this guide are the ones I have personally used and found valuable. But the ecosystem is vast and growing fast:

• Official MCP Servers — Reference implementations maintained by Anthropic
• awesome-mcp-servers — Community-curated list with 1,800+ servers
• MCP Registry — Searchable directory of published servers
• Smithery — MCP server marketplace with one-click install
• PulseMCP — Server discovery with reviews and ratings

New servers are being published daily. The categories growing fastest in 2026 are AI agent orchestration, enterprise SaaS integrations (Salesforce, HubSpot, Zendesk), and infrastructure-as-code tools.

Related Reading

• Claude Code MCP Setup Guide — Complete walkthrough of installing, configuring, and building MCP servers
• Building MCP Servers with TypeScript — Create your own MCP server from scratch
• MCP Protocol Explained — Technical deep dive into how MCP works under the hood
• MCP Security Guide — Best practices for running MCP servers safely
• Claude Code Complete Guide — Everything you need to know about Claude Code