In our guide to what MCP is, we explained how the Model Context Protocol lets AI systems connect to data and tools through a shared interface.
Now, let’s map the MCP ecosystem: servers, clients, and tools that are actively being developed for common AI workflows.
If you follow MCP, this is a strong time to explore what is available and decide which options fit your stack.
Let’s get started:
TLDR;
- AI Services
- Cloud Storage
- Development Tools
- Databases
- File Systems & Search
- Version Control & Project Management
- Communication & Social
- Cloud Platforms & Infra
- Web Automation & Search Engines
- Monitoring & Observability
- Finance & Payments
- Data Visualization
- Workflow & System Automation
- Note-Taking & Knowledge Bases
- Identity & Security
- Marketing & Analytics
- Aggregators & Connectors
- Virtualization & IoT
- Research & Literature
- Mapping & Location
1. AI Services
AI service MCP servers connect models to writing, image generation, forecasting, and other model-driven workflows without custom integration code
| Tool | What it does | Link |
| Agentset AI | Turns any private corpus into a Retrieval-Augmented-Generation (RAG) endpoint via MCP | https://github.com/agentset-ai/mcp-server |
| Hugging Face Spaces MCP | Chat with Spaces—images, audio, text—through standard MCP calls | https://github.com/evalstate/mcp-hfspace |
| Perplexity MCP | Bridges Anthropic/Claude clients to Perplexity’s answer engine | https://github.com/tanigami/mcp-server-perplexity |
| LlamaCloud MCP | Natural-language queries over managed vector indexes on LlamaCloud | https://github.com/run-llama/mcp-server-llamacloud |
| OpenAI MCP | Direct function-call wrapper for any OpenAI-compatible model | https://github.com/pierrebrunelle/mcp-server-openai |
| Any-Chat Completions MCP | Swap in Groq, xAI, or Perplexity endpoints transparently | https://github.com/pyroprompts/any-chat-completions-mcp |
| ZenML MCP | Inspect, trigger, and monitor MLOps pipelines inside chat | https://github.com/zenml-io/mcp-zenml |
| Chronulus AI MCP | Multimodal time-series forecasting agents for “predict anything” workflows | https://github.com/ChronulusAI/chronulus-mcp |
| PiAPI MCP | Generate Midjourney, Kling, Udio, Trellis media via one endpoint | https://github.com/apinetwork/piapi-mcp-server |
2. Cloud Storage
Model context protocol mcp servers that hook into Google Drive, OneDrive, even a supabase mcp bucket, letting any mcp client fetch or save files on the cloud.
| Tool | What it does | Link |
| Google Drive MCP | Read, search, and manage docs & folders with OAuth | https://github.com/modelcontextprotocol/servers-archived/tree/main/src/gdrive |
| VideoDB MCP (⭐) | Store, tag, and retrieve video moments with auto-transcription | https://github.com/video-db/agent-toolkit/tree/main/modelcontextprotocol |
| Microsoft 365 MCP | Unified access to OneDrive, Outlook mail, Excel, and Calendar | https://github.com/softeria/ms-365-mcp-server |
3. Development Tools
Devs get github mcp, figma mcp and a handy mcp python sdk; code, designs and tests move straight from IDE to chat with no copy-paste
| Tool | What it does | Link |
| OpenAPI Schema Explorer | Token-efficient browsing of huge API specs | https://github.com/kadykov/mcp-openapi-schema-explorer |
| Currents MCP (⭐) | Lets AI fix Playwright test failures reported to Currents | https://github.com/currents-dev/currents-mcp |
| marimo MCP | Interactive notebook-style coding via CodeMirror + MCP | https://github.com/marimo-team/codemirror-mcp |
| Comet Opik MCP | Query LLM telemetry & observability captured by Opik | https://github.com/comet-ml/opik-mcp |
| Figma MCP | Pull component measurements and color tokens right from your designs | https://github.com/GLips/Figma-Context-MCP |
| VSCode Devtools MCP | Exposes semantic IDE operations like “find usages” | https://github.com/biegehydra/BifrostMCP |
| Mastra MCP (⭐) | Chat with the entire Mastra.ai knowledge base | https://github.com/mastra-ai/mastra/tree/main/packages/mcp |
| Bucket MCP | Feature-flag management, data control, and rollout gating | https://github.com/bucketco/bucket-javascript-sdk/tree/main/packages/cli#model-context-protocol |
| EdgeOne Pages MCP | Deploy static HTML to Tencent EdgeOne and get a public URL | https://github.com/TencentEdgeOne/edgeone-pages-mcp |
4. Databases
From Postgres to Mongo, these third-party MCP servers store rows and vectors; some are self-hosted, some connect to AWS, and Supabase MCP offers a ready-made quick start
| Tool | What it does | Link |
| PostgreSQL MCP | Full-text SQL plus automatic schema inspection | https://github.com/modelcontextprotocol/servers-archived/tree/main/src/postgres |
| DuckDB MCP | OLAP queries on local files—great for analytics agents | https://github.com/ktanaka101/mcp-server-duckdb |
| Neon MCP (⭐) | Serverless Postgres with branch-per-feature workflow | https://github.com/neondatabase/mcp-server-neon |
| Redis MCP (⭐) | Key-value + vector store management through chat | https://github.com/redis/mcp-redis |
| Snowflake MCP | Enterprise cloud warehouse queries with access control | https://github.com/isaacwasserman/mcp-snowflake-server |
| MongoDB Lens MCP | Rich inspector for collections and indexes | https://github.com/furey/mongodb-lens |
| MySQL MCP | Classic relational data, supports read/write modes | https://github.com/designcomputer/mysql_mcp_server |
| Airtable MCP | Low-code tables, both CRUD and schema exploration | https://github.com/domdomegg/airtable-mcp-server |
| Qdrant MCP (⭐) | Vector search & memory recall for agents | https://github.com/qdrant/mcp-server-qdrant |
| BigQuery MCP #1 | TypeScript implementation for massive SQL datasets | https://github.com/LucasHild/mcp-server-bigquery |
| BigQuery MCP #2 | NPM package version—drop-in for Node apps | https://github.com/ergut/mcp-bigquery-server |
| Couchbase MCP (⭐) | Query KV & JSON docs with N1QL via MCP | https://github.com/Couchbase-Ecosystem/mcp-server-couchbase |
| TiDB MCP | Cloud-native MySQL-compatible HTAP database | https://github.com/c4pt0r/mcp-server-tidb |
| DBUtils MCP | Unified access layer bridging Postgres & SQLite | https://github.com/donghao1393/mcp-dbutils |
| NocoDB MCP | Chat-driven CMS built on NocoDB tables | https://github.com/edwinbernadus/nocodb-mcp-server |
| Excel MCP | Worksheet reading, charting, and formatting | https://github.com/haris-musa/excel-mcp-server |
| SQLite MCP | Zero-friction embedded database for local projects | https://github.com/modelcontextprotocol/servers-archived/tree/main/src/sqlite |
| VideoDB Director (⭐) | Create video workflows with database-backed retrieval | https://github.com/video-db/agent-toolkit/tree/main/modelcontextprotocol |
5. File Systems & Search
Local folders feel remote: an awesome mcp servers combo lets the AI scan disks, use everything search, and manage backup files, all through mcp github style calls
| Tool | What it does | Link |
| FileSystem (official) | Reference server for local disk operations | https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem |
| FileStash MCP | Access SFTP,S3,SMB,WebDAV,and more through one handler | https://github.com/mickael-kerjean/filestash/tree/master/server/plugin/plg_handler_mcp |
| Backup MCP | Snapshot & restore directories for agent workflows | https://github.com/hexitex/MCP-Backup-Server |
| llm-context.py | Share code snippets with LLMs via clipboard or MCP | https://github.com/cyberchitta/llm-context.py |
| Filesystem (Go) | Golang rewrite—lean and container-friendly | https://github.com/mark3labs/mcp-filesystem-server |
6. Version Control & Project Management
Ship code faster: github mcp server, gitlab, and linear tools let the model open pull requests or file bugs right inside the chat
| Tool | What it does | Link |
| GitHub MCP | Manage PRs, issues, labels, and reviews via chat | https://github.com/github/github-mcp-server |
| GitLab MCP | Control pipelines, merge requests, and epics | https://github.com/modelcontextprotocol/servers-archived/tree/main/src/gitlab |
| Git MCP | Local repo analysis, blame, diff, and log surfacing | https://github.com/modelcontextprotocol/servers/tree/main/src/git |
| Linear MCP | Create, triage, and update tickets from an LLM | https://github.com/jerhadf/linear-mcp-server |
| Gitingest MCP | Auto-summaries of GitHub repos for prompt injection | https://github.com/puravparab/Gitingest-MCP |
7. Communication & Social
Slack, TikTok, YouTube, even cursor mcp for desktop dev chat; the AI can read, post, or summarise any thread with one mcp cursor link.
| Tool | What it does | Link |
| Slack MCP | Search channels, post messages, build digests | https://github.com/zencoderai/slack-mcp-server |
| LINE OA MCP | Bring an AI agent into LINE Official Accounts | https://github.com/line/line-bot-mcp-server |
| Atlassian MCP | Unified Confluence + Jira operations | https://github.com/sooperset/mcp-atlassian |
| Bluesky MCP | Query federated feeds and threads | https://github.com/keturiosakys/bluesky-context-server |
| TikTok MCP | Fetch video metadata, captions, and stats | https://github.com/Seym0n/tiktok-mcp |
| Spotify MCP | Control playback, playlists, and library items | https://github.com/varunneal/spotify-mcp |
| YouTube MCP (yt-dlp) | Download subtitles & analyze transcripts | https://github.com/anaisbetts/mcp-youtube |
| YouTube Transcripts MCP | Alternative transcript-only implementation | https://github.com/kimtaeyoon83/mcp-server-youtube-transcript |
| WhatsMCP | Deploy an MCP server for WhatsApp via WAssist | https://wassist.app/mcp/ |
8. Cloud Platforms & Infra
Control Cloudflare, Kubernetes, or edge workers using mcp protocol; pass mcp certification (the mcp-1 badge) and run safer ops from chat
| Tool | What it does | Link |
| Cloudflare MCP (⭐) | Operate Workers, KV, R2, and D1 databases | https://github.com/cloudflare/mcp-server-cloudflare |
| Kubernetes MCP (Go) | Scale deployments, get pod logs, exec into containers | https://github.com/strowk/mcp-k8s-go |
| Tinybird MCP (⭐) | Run real-time SQL pipes and publish endpoints | https://github.com/tinybirdco/mcp-tinybird |
| Weibaohui K8M | Multi-cluster K8s management with 50 DevOps tools | https://github.com/weibaohui/k8m |
| Metoro MCP | Chat-query Kubernetes observability metrics | https://github.com/metoro-io/metoro-mcp-server |
9. Web Automation & Search Engines
Browser mcp and playwright mcp server let the bot click, scroll and scrape; zapier mcp or n8n mcp then push the data anywhere
| Tool | What it does | Link |
| Playwright MCP | Full browser automation plus scraping | https://github.com/executeautomation/mcp-playwright |
| Brave Search MCP | Privacy-first web results via API | https://github.com/brave/brave-search-mcp-server |
| Tavily MCP | Real-time answer API with citations | https://github.com/Tomatio13/mcp-server-tavily |
| Apify Actors MCP | Tap into 4,000+ scraping actors out-of-the-box | https://github.com/apify/actors-mcp-server |
| RAG Web Browser MCP | Search the web, scrape pages, return Markdown summaries | https://github.com/apify/mcp-server-rag-web-browser |
| Exa Search MCP (⭐) | Vector-based semantic search for the live web | https://github.com/exa-labs/exa-mcp-server |
| Puppeteer MCP | JS-controlled Chromium browsing via MCP | https://github.com/modelcontextprotocol/servers-archived/tree/main/src/puppeteer |
| Fetch MCP | Fetch and clean raw HTML or JSON payloads | https://github.com/modelcontextprotocol/servers/tree/main/src/fetch |
| Scrapeless MCP | Google SERP results through a single MCP endpoint | https://github.com/scrapeless-ai/scrapeless-mcp-server |
10. Monitoring & Observability
When something crashes the sequential thinking mcp flow shows Sentry or Raygun errors in plain chat and suggests a fix.
| Tool | What it does | Link |
| Sentry MCP | Query errors, performance data, and releases | https://github.com/modelcontextprotocol/servers-archived/tree/main/src/sentry |
| Raygun MCP | Access crash reports and Real User Monitoring stats | https://github.com/MindscapeHQ/mcp-server-raygun |
| Metoro MCP (duplicate entry above) | Kubernetes environment insights | https://github.com/metoro-io/metoro-mcp-server |
11. Finance & Payments
Stripe, PayPal, crypto prices—mcp agents handle money events, send invoices, or check wallet balances on demand
| Tool | What it does | Link |
| Stripe MCP (⭐) | Create charges, customers, and webhooks from chat | https://github.com/stripe/agent-toolkit/tree/main |
| PayPal MCP (⭐) | Payments, subscriptions, and transaction lookups | https://github.com/paypal/paypal-mcp-server |
| CoinMarket MCP | Real-time crypto prices from CoinMarketCap | https://github.com/anjor/coinmarket-mcp-server |
| Chargebee MCP (⭐) | SaaS subscription billing operations | https://github.com/chargebee/agentkit/tree/main/modelcontextprotocol |
| LongPort OpenAPI MCP (⭐) | Stock market data and trading via LongPort | https://github.com/longportapp/openapi |
| ZBD MCP (⭐) | Lightning-fast Bitcoin payments processing | https://github.com/zebedeeio/zbd-payments-typescript-sdk/tree/main/packages/mcp-server |
12. Data Visualization
The AI turns raw numbers into charts; one mcp llm request returns a clean Vega-Lite graph ready for reports
| Tool | What it does | Link |
| VegaLite MCP | Generate JSON specs and render charts | https://github.com/isaacwasserman/mcp-vegalite-server |
| AntV Chart MCP | Produce rich interactive visuals (G2/G6) | https://github.com/antvis/mcp-server-chart |
13. Workflow & System Automation
Make.com, Apple Shortcuts, or a simple shell: an mcp desktop host can fire scripts and close repetitive tasks while you watch
| Tool | What it does | Link |
| Make MCP (⭐) | Run Make.com scenarios directly from an LLM | https://github.com/integromat/make-mcp-server |
| Shell MCP | Secure shell access and Mac automation | https://github.com/rusiaaman/wcgw |
| Windows CLI MCP | Controlled PowerShell/CMD/Git Bash commands | https://github.com/SimonB97/win-cli-mcp-server |
| Apple Shortcuts MCP | Trigger iOS/macOS shortcut flows via chat | https://github.com/recursechat/mcp-server-apple-shortcuts |
| Windows Control MCP | Mouse, keyboard, and window management through NutJS | https://github.com/Cheffromspace/nutjs-windows-control |
| Command Line MCP | Execute any CMD command (use with care) | https://github.com/phialsbasement/cmd-mcp-server |
14. Note-Taking & Knowledge Bases
Notion mcp, Obsidian, Apple Notes; your private thoughts become searchable context for every chat.
| Tool | What it does | Link |
| Obsidian MCP #1 | Read & search Markdown vaults | https://github.com/MarkusPfundstein/mcp-obsidian |
| Notion MCP #1 | Create pages, databases, and tasks | https://github.com/danhilse/notion_mcp |
| Notion MCP #2 | Second implementation, Node-centric | https://github.com/suekou/mcp-notion-server |
| Apple Notes MCP | Read local macOS notes | https://github.com/sirmews/apple-notes-mcp |
| Todoist MCP | Manage personal tasks from chat | https://github.com/abhiz123/todoist-mcp-server |
| Google Keep MCP | CRUD operations on Keep notes | https://github.com/feuerdev/keep-mcp |
| Apple Books MCP | Turn EPUB/PDF libraries into a queryable KB | https://github.com/vgnshiyer/apple-books-mcp |
| Ebook-mcp | Local e-book browsing and summarization | https://github.com/onebirdrocks/ebook-mcp |
15. Identity & Security
Keycloak user tweaks, Semgrep scans—keep your model context protocol workspace safe without leaving the console
| Tool | What it does | Link |
| Keycloak MCP | User, group, and realm administration via MCP | https://github.com/ChristophEnglisch/keycloak-model-context-protocol |
| Semgrep MCP | Static code analysis and vuln surfacing | https://github.com/semgrep/mcp |
| Netwrix MCP (⭐) | Access Analyzer for permission audits | https://github.com/netwrix/mcp-server-naa |
16. Marketing & Analytics
Facebook Ads stats, Fathom web traffic, plus AI copy helpers: perfect when ai mcp needs to boost SEO or track spend
| Tool | What it does | Link |
| Open Strategy Partners Tools | Standardized editing codes & metadata generators | https://github.com/open-strategy-partners/osp_marketing_tools |
| Fathom Analytics MCP | Privacy-focused site analytics reporting | https://github.com/mackenly/mcp-fathom-analytics |
| Facebook Ads MCP | Programmatic ad management & reporting | https://github.com/gomarble-ai/facebook-ads-mcp-server |
17. Aggregators & Connectors
One hub, many apps: pipedream, zapier mcp, and toolhive install or route hundreds of integrations with one command
| Tool | What it does | Link |
| Pipedream MCP | 8,000+ APIs callable through one server | https://github.com/PipedreamHQ/pipedream/tree/master/modelcontextprotocol |
| Zapier MCP | Zapier’s 8 k app catalog via MCP | https://zapier.com/mcp |
| Remote MCP | Centralised management for distributed servers | https://github.com/ssut/Remote-MCP |
| yamcp | Workspace manager to scan/monitor local MCP servers | https://github.com/hamidra/yamcp |
| ToolHive | Containerized deployment helper for MCP servers | https://github.com/StacklokLabs/toolhive |
| mcp-get | CLI installer/upgrader for NPM-based servers | https://github.com/michaellatman/mcp-get |
18. Virtualization & IoT
Spin up test boxes in E2B or talk to smart sensors via Coreflux MQTT—all through the same mcp server list
| Tool | What it does | Link |
| E2B MCP (⭐) | Spin up disposable dev containers in the cloud | https://github.com/e2b-dev/mcp-server |
| Docker MCP | Manage local Docker daemon and compose stacks | https://github.com/QuantGeekDev/docker-mcp |
| Coreflux MQTT MCP | Turn MQTT broker into an automation hub for IoT | https://github.com/CorefluxCommunity/CorefluxMCPServer |
19. Research & Literature
Search papers on arXiv, books on Open Library, or code on PapersWithCode; a single what is mcp in ai query can now cite real sources
| Tool | What it does | Link |
| ArXiv MCP | Search and fetch research papers | https://github.com/blazickjp/arxiv-mcp-server |
| PapersWithCode MCP | Track SOTA, conferences, and code bases | https://github.com/hbg/mcp-paperswithcode |
| Open Library MCP | Look up book & author info via Internet Archive | https://github.com/8enSmith/mcp-open-library |
| Ancestry MCP | Parse GEDCOM files and genetic datasets | https://github.com/reeeeemo/ancestry-mcp |
20. Visualization, Mapping & Location
Google Maps layers or QGIS projects; ask “Where’s the site?” and the model context protocol ai stack returns routes and geo shapes
| Tool | What it does | Link |
| Google Maps MCP | Geocoding, routing, and place details | https://github.com/modelcontextprotocol/servers-archived/tree/main/src/google-maps |
| QGIS MCP | Operate QGIS Desktop layers and projects | https://github.com/jjsantos01/qgis_mcp |
21. Miscellaneous & Fun
Unity game hooks, mcp router tricks, and blender mcp experiments show how far an awesome mcp setup can stretch
| Tool | What it does | Link |
| Unity MCP | Integrate AI into Unity Editor workflows | https://github.com/IvanMurzak/Unity-MCP |
| Unity3d Game MCP | Alternative Unity server for game engines | https://github.com/CoderGamester/mcp-unity |
| Model Context Protocol Open Library | Access classic lit with “awesome MCP” flair | (same as Open Library MCP above) |
| MCP Router | Gateway that auto-routes calls to sub-servers | https://github.com/mcp-router/mcp-router |
Sources: Awesome MCP Servers by appcypher and Awesome MCP Servers by punkpeye
Frequently Asked Questions
What is an MCP server in AI?
An MCP server is the layer that exposes a tool or data source to an AI through a standard interface. It lets one MCP setup connect to cloud storage, databases, search, or version-control tools without a separate custom integration for each one.
What is the difference between an MCP server and an MCP client?
The server provides the capability, and the client uses it. If a tool exposes Google Drive, GitHub, or a database over MCP, that tool is the server; the chat app, assistant, or development environment making the request is the client.
What are MCP servers actually useful for?
MCP servers let an assistant reach the right systems during a conversation instead of relying only on its base model. Common uses include searching documents, reading cloud files, querying databases, using development tools, and automating support or workflow actions.
Does MCP work with ChatGPT?
For ChatGPT specifically, a blanket yes is not verified. If you want to use ChatGPT with MCP, verify whether your client or bridge supports MCP tool calls before rollout. What is established is the broader pattern: MCP connects AI systems to tools and data through a standard interface.
Should you use a hosted MCP server or build your own?
If your priority is faster setup and less implementation work, a hosted or no-code MCP option is usually the better fit. Building your own is more reasonable when existing servers and connectors do not cover the systems or workflows you need.
Are MCP servers secure enough for business data?
MCP servers can be secure enough for business data, but you should verify controls before connecting sensitive systems. Check for authenticated access, least-privilege permissions, logs or audit trails, and clear data-handling terms. Useful trust signals include SOC 2 Type 2 certification, GDPR compliance, and a policy that customer data is not used for model training.
Related Resource:
- Source-grounded Trae agents See how to reduce generic Trae responses by connecting agents to CustomGPT.ai over MCP.

Priyansh is a Developer Relations Advocate at CustomGPT.ai who writes deeply researched technical content on RAG APIs, AI agent development, and cloud-native tools.