The Claude Code MCP Servers I Actually Use Every Day

The Claude Code MCP Servers I Actually Use Every Day

ByKjetil Furas Updated on

I run 9 MCP servers in my Claude Code setup. Not because I collected them — because each one replaced a manual step in a workflow I run daily. After three months of daily usage, these are the Claude Code MCP servers that actually earned their place.

Most “best MCP servers” lists rank by GitHub stars. This ranks by how often I actually use each one and what it replaced.

What MCP servers do in Claude Code

MCP (Model Context Protocol) gives Claude Code access to external tools and APIs through a standardized interface. Instead of writing custom scripts for every integration, you install an MCP server and Claude gets new capabilities — posting tweets, querying databases, managing files in cloud storage.

Each server exposes tools that Claude can call directly. Install the GitHub MCP server and Claude can create PRs, review code, and manage issues without you touching the GitHub UI. Install the Notion MCP server and Claude can search, create, and update pages.

The setup lives in your Claude Code settings file. Point it at the server, configure credentials, and the tools appear in Claude’s toolkit immediately.

The MCP servers I use daily

1. Context7 — Live documentation lookup

{
  "mcpServers": {
    "context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp"]
    }
  }
}

Context7 fetches current documentation for any library or framework. When Claude needs to check the latest API syntax for a package, it pulls the actual docs instead of relying on training data that might be months behind.

I use this dozens of times per day. Every time I work with a library I haven’t touched in a while — Playwright, FFmpeg flags, YouTube API endpoints — Context7 gives Claude the current docs in seconds.

What it replaced: Manually searching docs, copying snippets into chat, correcting Claude when its training data was outdated.

2. YouTube Data + Analytics API

My YouTube MCP server wraps the Data API, Analytics API, and Reporting API into one server. Claude can upload videos, pull analytics, read comments, manage playlists, and check traffic sources — all through natural language.

# Example: Pull top performing Shorts
mcp__youtube__youtube_analytics_top_shorts(days=28, max_results=20)

I run analytics reviews weekly. Instead of logging into YouTube Studio, clicking through tabs, and cross-referencing metrics manually, I ask Claude to pull the data and identify patterns. It takes 30 seconds instead of 15 minutes.

What it replaced: YouTube Studio dashboard, manual spreadsheet tracking, switching between Analytics tabs.

3. X (Twitter) API

Posts tweets, reads my timeline, deletes posts, generates images, and scans for viral content in my niche. The MCP server wraps both the v2 API for posting and Playwright for scraping engagement data that the API doesn’t expose.

# Post with optional media
mcp__x-mcp__post_tweet(text="...", image_path="/path/to/media.mp4")

# Pull my recent tweets with metrics
mcp__x-mcp__get_my_tweets(count=20)

I post 3-5 times per week through this server. Claude drafts the post following my brand voice rules, I approve, and it publishes. The viral tweet scanner runs weekly to find trending content in the AI/automation niche.

What it replaced: Opening Twitter, composing manually, checking analytics in a separate tab, manually searching for trending content.

4. Bluesky

Same content pipeline as X, adapted for Bluesky’s 300 character limit and different link preview behavior. Simple MCP server — create posts, get timeline, manage profile.

mcp__bluesky-mcp__bluesky_create_post(text="...")

Cross-posting takes 10 seconds. Claude adapts the X post for Bluesky’s format and publishes. No manual reformatting.

What it replaced: Copy-pasting from X to Bluesky, manually trimming for the 300 char limit, reformatting links.

5. Google Search Console

Pulls search analytics, requests indexing, submits sitemaps, and inspects URLs. Essential for content strategy — I use it to find keyword opportunities before writing blog posts.

# Find keywords with high impressions but low CTR
mcp__gsc__gsc_search_analytics(
  site_url="https://kjetilfuras.com/",
  days=28,
  dimensions="query",
  row_limit=25
)

Every blog post starts with GSC data. Claude pulls the keywords, identifies where I have impressions but low click-through rates, and picks a target keyword with ranking potential. This is how I choose what to write about.

What it replaced: Logging into Google Search Console, exporting CSVs, analyzing in spreadsheets.

6. Airtable

Full CRUD on Airtable bases — create records, update fields, search, and manage tables. I use it as the backend for tracking Skool community members, content outcomes, and task status.

# Search for a specific member
mcp__airtable__search_records(
  base_id="appIxRJJ7JhlJ5tht",
  table_id="tblsEH6vkHtcT1d47",
  search_term="[email protected]"
)

My Skool member sync runs through this — Playwright exports members from Skool, diffs against the Airtable base, and upserts new/churned members automatically.

What it replaced: Manual Airtable updates, copy-pasting member data, weekly manual audits.

7. Notipo (Notion → WordPress publishing)

Notipo is my own SaaS tool that syncs Notion pages to WordPress automatically. The MCP server lets Claude create blog posts, set SEO metadata, trigger publishes, and check sync job status — all without leaving the terminal.

# Create and publish a blog post
notipo posts create \
  --title "Post Title" \
  --body "$(cat draft.md)" \
  --category "AI" \
  --seo-keyword "target keyword" \
  --slug "post-slug" \
  --publish --wait

Every blog post on kjetilfuras.com goes through this pipeline. Claude drafts the post, saves it as markdown, publishes via Notipo, and the content lands on WordPress with SEO metadata, categories, and tags in one command.

What it replaced: Manual WordPress editing, copy-pasting from Notion, setting SEO fields by hand in Yoast/RankMath.

8. Gmail + Google Calendar

Read emails, send messages, manage calendar events, search across accounts. I use this for checking if important emails came in and scheduling around existing calendar blocks.

# Search for recent emails about a topic
mcp__gmail__gmail_search(query="from:kit.com subject:broadcast", max_results=5)

What it replaced: Switching to Gmail, searching manually, copying relevant info back into the conversation.

9. Content tools (custom)

A custom MCP server that wraps my own scripts — image generation via Gemini API, topic research, reaction clip search, thumbnail generation, and text-to-speech for video narration.

# Generate a thumbnail
mcp__content-tools__generate_thumbnail(
  title="MAIN TEXT",
  subtitle="optional",
  output_path="/path/to/thumbnail.jpg"
)

# Research trending topics
mcp__content-tools__research_topics(query="claude code", days=7, limit=10)

This is the server that ties everything together. It wraps scripts I would otherwise run manually from the terminal into tools Claude can call mid-conversation.

What it replaced: Running Python scripts manually, remembering flag syntax, copying output back into the conversation.

How to decide which MCP servers you need

Don’t install servers because they’re popular. Install them because you do the task manually more than twice a week.

My process:

  1. Track your manual steps for a week. Every time you leave Claude Code to do something in another app — check email, post a tweet, look up analytics — write it down.
  2. Check if an MCP server exists for that app. The MCP server directory has hundreds. Context7, Playwright, and the major SaaS integrations cover most use cases.
  3. If no server exists, wrap your script. The FastMCP framework lets you turn any Python function into an MCP tool in under 20 lines. My content-tools server started as one script and grew to six tools.
  4. Give it two weeks. Some servers sound useful but you never call them. Others become indispensable on day two. Remove what you don’t use — every server adds startup time.

The setup that matters more than which servers you pick

The servers are just plumbing. What makes them useful is the automation layer on top:

  • Skills — Markdown files that tell Claude when and how to use each tool. My Claude Code agent skills post covers this in detail.
  • Hooks — Shell commands that fire on Claude Code lifecycle events. Auto-format code after edits, run tests after file changes, block dangerous commands.
  • CLAUDE.md — Your global system prompt. It tells Claude which MCP tools to prefer for which tasks, so it doesn’t pick the wrong server when multiple ones could work.

The servers give Claude hands. The skills tell it what to build.

FAQ

How many MCP servers can Claude Code run at once?

There is no hard limit, but each server adds to startup time and memory usage. I run 10 comfortably. Beyond 15, you will notice slower session starts. Only keep servers you actively use.

Do MCP servers cost anything?

The servers themselves are free and open source. The APIs they connect to may have costs — YouTube API has quotas, X API has rate limits, Airtable has row limits on free plans. The MCP layer adds zero cost on top.

Can I build my own MCP server?

Yes. FastMCP (Python) and the official TypeScript SDK make it straightforward. If you have a Python script that does something useful, you can wrap it as an MCP tool in 15 minutes. My content-tools server is a single Python file with six functions decorated as tools.

Are MCP servers secure?

Each server runs as a separate process with its own credentials. Claude can only call tools that the server exposes — it cannot access the underlying system beyond what you configure. Keep API keys in environment variables, not in the server config file.

I’m documenting the full MCP server setup — every config file, every custom server, the automation layer that ties them together — in my Build & Automate community. Real production configs, not demo projects.

This post was published using Notipo — my Notion-to-WordPress sync tool. Write in Notion, publish to WordPress automatically.

Kjetil Furas

Kjetil, an IT expert from Norway, combines over a decade of tech experience with AI expertise. He writes clear, insightful content to guide readers through complex technological topics.

Similar Posts