The only list you need. Each server in 1-2 lines with a link.
There are over 10,000 MCP servers listed across directories right now. Most of them are weekend projects that break the first time you try them.
I tested dozens over last year of vibecoding and kept the ones that actually work, are actively maintained, and solve a real problem.
These are the MCP servers worth installing, sorted by what they do.
If you don't know what MCP is: it's how you connect Claude to external tools.
Before we dive in, I share daily notes on AI & vibe coding in my Telegram channel: https://t.me/zodchixquant🧠
Search & Research 🔍
01 — Tavily — AI-optimized web search. Returns clean content, not just links. Built specifically for AI agents. Free: 1,000 queries/mo
https://github.com/tavily-ai/tavily-mcp
02 — Exa — Semantic search. Finds pages by meaning, not keywords. Great for research and finding similar content. Free: 1,000 searches/mo
https://github.com/exa-labs/exa-mcp-server
03 — Brave Search — Independent web search index, no Google tracking. 6 tools: web, local, image, video, news, summarizer. Paid: from $5/mo
https://github.com/anthropics/anthropic-quickstarts
04 — Perplexity — Answer-engine with deep research and reasoning. Returns synthesized answers, not raw results. Free tier available
https://github.com/ppl-ai/perplexity-mcp
05 — Context7 — Live documentation for any framework. Claude stops hallucinating outdated APIs. Essential for devs working with fast-moving libraries. Free & open-source
https://github.com/context-labs/context7
Web Scraping & Data Extraction 🌐
06 — Firecrawl — Any URL to clean markdown in seconds. Autonomous research agent, JS rendering, full site crawls. The go-to for RAG pipelines. Free: 500 lifetime credits
https://github.com/firecrawl/firecrawl-mcp-server
07 — Apify — 3,000+ ready-made scrapers. Google Maps, Amazon, LinkedIn, social media. If a website exists, Apify probably has a scraper for it. Free: $5/mo credits
https://github.com/apify/actors-mcp-server
08 — Bright Data — Heavy anti-bot bypass. Cloudflare, CAPTCHA, enterprise protection. When Firecrawl and Apify can't get through. Paid only
https://github.com/nicholasgriffintn/bright-data-mcp
09 — Crawl4AI — Free open-source crawling with the best markdown extraction. 61k+ GitHub stars. No API key needed. Free & open-source
https://github.com/unclecode/crawl4ai
Browser Automation 🖥️
10 — Playwright — Claude controls a real Chrome browser. Clicks, fills forms, navigates, takes screenshots. Testing without writing test scripts. Free & open-source
https://github.com/anthropics/anthropic-quickstarts
11 — Browserbase — Cloud-hosted browser for AI. No local Puppeteer setup. Each session returns a live browser URL your agent controls. Free: 100 sessions/mo
https://github.com/browserbase/mcp-server-browserbase
Dev Tools & Code 💻
12 — GitHub — PRs, issues, code search, CI/CD workflows. The first MCP server every developer should install. Free with GitHub account
https://github.com/github/github-mcp-server
13 — Linear — Issue tracking without leaving Claude. Create tickets, search with filters, update status, manage sprints. Free tier available
https://github.com/linear/linear-mcp-server
14 — Sentry — Production errors in context. Full stack traces, breadcrumbs, crash diagnostics. From alert to fix without switching tools. Free: 5K errors/mo
https://mcp.sentry.dev
15 — Vercel — Deploy, check build logs, inspect errors, manage projects. Debug failed deployments in one loop. Free tier available
https://github.com/vercel/mcp-server-vercel
16 — Jira / Atlassian — JQL search, ticket creation, sprint management, workflow transitions. For teams that live in Jira. Free: up to 10 users
https://github.com/atlassian/mcp-server-atlassian
Databases 🗄️
17 — Supabase — Postgres + auth + storage through prompts. Built for dev environments. Read-only mode available. Free tier available
https://github.com/supabase/mcp-server-supabase
18 — PostgreSQL — Natural language database queries. Schema inspection, data analysis, migration help. Free & open-source
https://github.com/anthropics/anthropic-quickstarts
19 — MongoDB — 40+ tools. Atlas management, performance advisors, aggregation pipelines. The most comprehensive database MCP. Free: 512MB cluster
https://github.com/mongodb/mongodb-mcp-server
20 — Neo4j — Graph database queries. Relationship analysis, knowledge graphs, connected data exploration. Free: community edition
https://github.com/neo4j/neo4j-mcp-server
Vector & AI Memory 🧠
21 — Pinecone — Cloud vector search with reranking. For RAG pipelines and semantic retrieval at scale. Free: 2GB storage
https://github.com/pinecone-io/pinecone-mcp
22 — Qdrant — Semantic memory and vector search. Open-source, self-hostable. Free & open-source
https://github.com/qdrant/mcp-server-qdrant
23 — Chroma — Lightweight vector database. Good for local dev and prototyping RAG systems. Free & open-source
https://github.com/chroma-core/chroma-mcp
24 — Memory MCP — Knowledge graph-based persistent memory. Claude remembers across sessions without Projects. Free & open-source
https://github.com/anthropics/anthropic-quickstarts
Productivity & Collaboration 📋
25 — Notion — Documents, wikis, databases through prompts. Search, summarize, and act on your workspace without opening Notion. Free tier available
https://github.com/anthropics/anthropic-quickstarts
26 — Slack — Read threads, search history, send messages. "Summarize what the team discussed about the launch" actually works. Free with Slack workspace
https://github.com/anthropics/anthropic-quickstarts
27 — Todoist — Task management through Claude. Create, organize, and track tasks without opening the app. Free tier available
https://github.com/abhiz123/todoist-mcp-server
28 — Zapier — Automation between any apps. Trigger workflows across Slack, Gmail, Trello, Sheets, and 6,000+ other apps from one prompt. Free: 100 tasks/mo
https://github.com/zapier/zapier-mcp-server
Business & Finance 💰
29 — Stripe — Payments, subscriptions, invoices, customer data through prompts. Check revenue, manage subscriptions, debug payment issues. Free: test mode
https://github.com/stripe/agent-toolkit
30 — HubSpot — CRM through Claude. Contacts, deals, pipeline management, activity tracking without the HubSpot UI. Free CRM available
https://github.com/hubspot/mcp-server-hubspot
Design & Media 🎨
31 — Figma — Design files to code. Read tokens, inspect components, grab screenshots of frames. The design-to-code gap shrinks to zero. Free with Figma account
https://github.com/nicholasgriffintn/figma-mcp
32 — Bannerbear — Automated image and banner generation from templates. Social media graphics, OG images, certificates at scale. Free: 30 images/mo
https://github.com/bannerbear/bannerbear-mcp
Infrastructure & Monitoring ⚙️
33 — Cloudflare — Manage Workers, KV storage, D1 databases, DNS, and security rules through Claude. Free tier available
https://github.com/cloudflare/mcp-server-cloudflare
34 — Docker — Container management through prompts. List, start, stop, inspect containers and images. Free & open-source
https://github.com/docker/docker-mcp
35 — Grafana — Query dashboards, check metrics, investigate alerts. Monitoring without tab-switching. Free: Grafana Cloud
https://github.com/grafana/mcp-grafana
How to install any MCP server
All servers use the same pattern:
For Claude Desktop, add to your config file:
Where to start
You don't need all 35.
Start with 3-5 that solve problems you have right now:
Each MCP server you add uses token context. 3-5 servers is the sweet spot. More than that and you're burning tokens on tool descriptions before you even ask a question.
Claude Code has a Tool Search feature that lazy-loads servers to reduce this, but keep it lean.
I share daily notes on AI, finance, and vibe coding in my Telegram channel: https://t.me/zodchixquant
Thanks for reading 🙏🏼
