Top 100 Mcp Servers

The most-installed MCP server. Secure local file read/write with configurable allowlist paths. Reference implementation in modelcontextprotocol/servers and the default first-server any client pairs with.

Anthropic-published server giving agents full repo operations — read/write files, open issues, create PRs, search code, manage workflows. The reference enterprise dev server.

Read-only Postgres connector with schema inspection. Most-used database MCP server; the archived reference is the upstream for many forks adding write capability.

Channel management, messaging, and search across Slack workspaces. Lets agents read threads, post messages, and react — table-stakes for any internal-tooling agent.

Persistent knowledge-graph memory server. Stores typed entities and relations across sessions; the canonical 'remember things between conversations' server.

Structured chain-of-thought server. Lets agents externalise multi-step reasoning into revisable thought sequences. Surprisingly impactful on complex coding tasks.

HTTP fetch and HTML→markdown extraction in one server. Lighter-weight than browser automation when all you need is to grab a page.

Local git operations — read commits, search blame, diff branches, stage changes. Often paired with Filesystem for full repo-aware agents.

Web and local search via the Brave Search API. The default web-search MCP server for agents that need privacy-respecting results.

Browser automation server — navigate, screenshot, click, evaluate JS. Default for scraping and end-to-end testing inside agent workflows.

Microsoft-maintained Playwright integration with first-class accessibility-tree navigation. Superior to Puppeteer for agents that need semantic page understanding.

Connects agents to 7,000+ app actions through a single unified server. The 'one server for everything' option for non-technical users.

Read and write Notion pages, databases, and blocks. Workspace management and content publishing for agents that live inside Notion-centric orgs.

Full Linear integration — read/write issues, projects, cycles, comments. Remote MCP endpoint launched 2026; the canonical 'agentic project management' server.

Official remote MCP for Jira + Confluence. Lets agents query tickets, transition states, and edit Confluence pages directly.

Retrieve and analyse Sentry issues, breadcrumbs, and releases. Standard for 'find the error and fix it' agentic loops.

Built for developer-centric orgs — turns Supabase dev workflows into natural-language commands with safe schema management. Project create, migrate, branch.

Deployment, env-var, and project management on Vercel. The 'ship this to production' tool for agentic IDEs.

Manage Cloudflare zones, Workers, R2, KV, and D1 from an agent. Strong on infra-as-code-from-natural-language workflows.

Serverless Postgres provider with first-class MCP. Branch databases for agents, then merge or drop. Differentiates by per-agent ephemeral DB pattern.

CRM data access for agents — contacts, companies, deals, tickets. Remote endpoint launched 2026; popular for sales-ops automation.

Stripe API access for agents — payments, customers, subscriptions, products. The reference billing-platform MCP server.

Managed vector database with MCP wrapper. Agents recall semantically similar data points and write back upserts.

Open-source vector database powering context memory across agent frameworks. The community-favorite for self-hosted RAG.

Vector + keyword hybrid search via MCP. Common backend for production agentic RAG with mixed structured/unstructured data.

Lightweight open-source vector store with MCP integration. The 'just works' choice for local agent memory and prototyping.

Hosted RAG-as-a-service that excels at searching vast document sets with API-standardized prompts. Strong semantic ranking out of the box.

Local SQLite querying and schema introspection. The simplest possible database server — perfect for single-user agent state.

Official MongoDB MCP — find, aggregate, insert across collections with full schema inspection. Drives many agent-backed search/analytics pipelines.

MySQL/MariaDB connector with schema inspection and parameterised query support. Common substitute for the Postgres server in MySQL-shop environments.

Query Snowflake warehouses, manage schemas, and inspect lineage. The reference enterprise-data-warehouse MCP server.

Google BigQuery MCP for agentic data analysis. Query, manage datasets, and run jobs from inside Claude/Cursor.

Databricks integration covering SQL warehouses, Unity Catalog, and Genie. Bridges the gap between Claude and the lakehouse.

Search and read files across Drive. Often the simplest way to give an agent access to organisational documents.

Read and create calendar events. Standard wiring for scheduling agents and meeting-prep loops.

Gmail message search, draft, and label management. The reference email server for Workspace-based agents.

Discord channel and DM messaging. Popular in community-management workflows and bot-replacement agents.

Telegram bot and user-mode messaging via MCP. Common in remote alerting and personal-assistant agent setups.

GitLab equivalent of the GitHub server — repos, issues, MRs, pipelines. Default for self-hosted GitLab environments.

Bitbucket Cloud and Data Center integration — repos, pull requests, pipelines, comments. The Atlassian-shop counterpart to GitHub MCP.

AWS Labs' suite of MCP servers covering Lambda, S3, DynamoDB, EC2 and more. The reference cloud-vendor MCP collection.

Microsoft's Azure MCP family — Cosmos DB, Storage, OpenAI, and the Azure Resource Manager. Tight integration with VS Code agents.

Google Cloud Platform MCP — Compute Engine, Cloud Storage, GKE, and Vertex AI. The third-party alternative to AWS/Azure servers.

Tavily search API wrapper — high-quality summarized web search optimised for AI consumption. Strong free tier.

Neural search engine designed for AI agents — semantic search with rich content extraction. Differentiates from keyword-based engines.

Perplexity-powered web research with citations. Useful when you need an answer-shaped result rather than ranked links.

Production-grade scraper with built-in conversion to markdown and JSON. Strong on JS-heavy SPAs and pagination loops.

AI-driven browser navigation server that wraps Playwright with autonomous task execution. Common in research-and-extract agents.

Apify Actors store with 5000+ pre-built scrapers and automations exposed through a single MCP server. Strong for one-off data jobs.

Bright Data's proxy + scraping infrastructure surfaced as an MCP. Heavy enterprise scraping at scale with built-in rotation.

Time-zone and date arithmetic server. Looks trivial but solves a recurring agent failure mode: getting 'today' wrong across zones.

Reference test server exercising every MCP primitive (tools, resources, prompts, sampling). Used for client conformance testing.

Sourcegraph code search across multiple repos with semantic + structural queries. Strong for refactors across large monorepos.

Docker daemon control — list, build, run, exec containers. Standard tooling for agents that need to test code in isolation.

kubectl-equivalent MCP for inspecting and mutating cluster state. Common in SRE-assist agents and incident-response loops.

Terraform plan/apply/state inspection via MCP. Lets agents generate, validate, and apply IaC changes with human-in-the-loop approval.

Redis key-value and stream operations. Common as a cache + ephemeral state store for multi-agent orchestration.

Elasticsearch query DSL exposed as MCP tools. The reference log/full-text-search server for ops agents.

ClickHouse OLAP queries via MCP. Strong fit for product-analytics agents over event data.

Embedded DuckDB OLAP for local CSV/Parquet exploration. The fastest 'just analyse this file' server in the ecosystem.

Airtable base read/write — records, views, schema. Heavy use in operations-team agents who already live in Airtable.

ElevenLabs voice generation MCP — text-to-speech, voice cloning, dubbing. The default audio-output server for talking agents.

Run any Replicate model from an agent. The 'one server, many models' option for ad-hoc image, video, and audio gen.

fal.ai's inference platform exposed as MCP. Strong for low-latency image and audio model calls inside agentic loops.

Hugging Face Hub access — model search, dataset listing, Spaces invocation. The reference open-AI-platform MCP.

Product analytics MCP — funnels, retention, feature flags, dashboards. Lets PMs and devs interrogate user behaviour conversationally.

Mixpanel event querying and dashboard creation via MCP. Used in growth-team agents that auto-generate analysis.

Twilio Segment customer-data-platform MCP. Manages sources, destinations, and tracking plans through an agent.

Search infra MCP — indexes, queries, A/B tests. Common in e-commerce agents that need to tune search relevance.

Lightweight, blazing-fast search engine with MCP wrapper. Strong open-source alternative to Algolia for self-hosted setups.

SMS, voice, and WhatsApp messaging through Twilio's MCP. The reference comms server for SMS-led agentic workflows.

Transactional email through Resend's developer-first API. Popular in agent-led marketing and onboarding flows.

Twilio SendGrid email-platform MCP — templates, contacts, campaigns. Default email MCP for traditional marketing stacks.

Banking and financial-account data via Plaid. The reference fintech MCP for personal-finance and lender-side agents.

Shopify admin API — products, orders, customers, inventory. Default for e-commerce-shop agentic operations and merchandising.

Asana project and task management — create, update, assign across workspaces. Common in PM agents that bridge Slack/Asana.

ClickUp task, list, and doc management via MCP. Heavy adoption in agency / consulting teams.

Boards, items, and automations on monday.com surfaced as MCP. Strong in operations-led agentic workflows.

Trello board and card automation via MCP. Lightweight choice for personal-productivity agents.

Personal knowledge management — read, write, search across Obsidian vaults. The dominant self-hosted notes MCP.

Atlassian Confluence integration for page CRUD, search, and labels. Bundled in the official Atlassian Remote MCP.

Community Jira MCP servers for users who can't deploy the official Atlassian remote. Common in pre-2026-MCP enterprise environments.

Spotify playback control and library management. Surprisingly popular as the 'fun' demo MCP that proves the protocol works.

YouTube Data API MCP — video metadata, comments, channel stats. Common in content-strategy and competitive-analysis agents.

Reddit listings, comments, and search via MCP. Useful in research agents and community-monitoring workflows.

Hacker News API MCP — top stories, comments, search. Lightweight feed source for tech-research agents.

ArXiv paper search and fetch via MCP. Default research-paper server for science-adjacent agentic workflows.

Wikipedia article retrieval and search. Common as a grounding source for agents that need encyclopedic context.

Maps search, geocoding, and directions via Google Maps Platform. Reference travel/logistics MCP.

Open alternative to Google Maps for geocoding, routing, and POI lookup. Used in privacy-sensitive geospatial agents.

Smithery's hosted MCP registry — install thousands of MCP servers with one CLI command across Claude Desktop, Cursor, Cline. Distribution layer for the ecosystem.

Community-maintained directory of curated MCP servers. Browse, search, and pull configs across hundreds of community servers.

Composio's 250+ pre-built integrations exposed as a single MCP. Competes with Zapier MCP as the 'one server for everything'.

Pulumi infrastructure-as-code MCP — generate, plan, and apply Pulumi stacks. Alternative to Terraform MCP for TypeScript/Python IaC shops.

AI image generation MCP from Everart — the official archived reference for adding image generation tools to clients.

AWS Bedrock Knowledge Base retrieval surfaced as MCP. Used in agentic RAG pipelines built on top of Bedrock-managed corpora.

Mem0 personal-memory layer with MCP wrapper. Provides 'this user said X last week' style long-term memory for assistants.

LlamaIndex's MCP-compatible server for agentic RAG over indexed document collections. Strong for codebase + docs hybrid retrieval.

Community MCP wrapping the NotebookLM CLI for notebook/source/asset management. Used in /vad-style daily-pipeline agents.

Official MCP debugger and protocol inspector. Lets you connect to any MCP server and exercise its tools/resources interactively — essential during development.