Best MCP servers — the directory

Read and write local files within allowed directories.

A knowledge-graph memory the model can read and write across turns.

A tool for step-by-step reasoning and problem decomposition.

A reference server exercising every MCP feature — useful for testing clients.

Fetch a URL and convert the page to markdown for the model.

Read, search, and manipulate a local Git repository.

Time and timezone conversions.

Query and inspect a local SQLite database.

Repositories, issues, and pull requests via the GitHub API.

Projects, issues, and merge requests via the GitLab API.

Talk to a running JetBrains IDE — files, refactors, and run configs.

Run code snippets in many languages and return the output.

Run code in a secure cloud sandbox.

Run terminal commands and edit files on your machine.

Read and drive the active iTerm terminal session (macOS).

Run allow-listed shell commands and return their output.

Pull up-to-date, version-specific library docs into the model.

Generate React UI components from a natural-language prompt.

Read Figma files and frames so an agent can implement the design.

Convert PDFs, Office docs, and more to markdown.

Drive a real browser for navigation, scraping, and testing.

Browser automation and screenshots via Puppeteer.

Cloud headless browsers for automation at scale.

Run Apify Actors to scrape and automate the web.

Web and local search via the Brave Search API.

Web search and extraction tuned for LLMs.

Neural web search and crawling for AI agents.

Web search via DuckDuckGo, no API key required.

Ask the Perplexity Sonar models for sourced answers.

Crawl, scrape, and search sites into clean markdown.

Search and fetch papers from arXiv.

Read-only SQL queries against a Postgres database.

Read and write keys in a Redis instance.

Query and manage a MongoDB database.

One SQL server for Postgres, MySQL, SQL Server, SQLite, and MariaDB.

Manage a Supabase project — tables, SQL, and edge config.

Manage Neon serverless Postgres projects and run SQL.

Read and write Airtable bases and records.

Search and inspect Elasticsearch indices.

Run analytical SQL against a ClickHouse database.

Query and manage Pinecone vector indexes.

Store and retrieve vectors from a Qdrant database.

Generate charts and graphs from data (AntV).

Retrieve from an AWS Bedrock knowledge base.

Manage Cloudflare Workers, KV, R2, and DNS.

Manage Heroku apps, dynos, and add-ons.

Query and manage Azure resources.

Inspect and operate a Kubernetes cluster via kubectl.

Read and update Notion pages and databases.

Read and search an Obsidian vault.

Manage Linear issues, projects, and cycles.

Manage ClickUp tasks, lists, and docs.

Manage Trello boards, lists, and cards.

Jira issues and Confluence pages.

Search and read files in Google Drive.

Read, search, and send Gmail messages.

Post and read messages in Slack channels.

Send and read messages in Discord servers.

Send SMS and use the Twilio APIs.

Generate images with the EverArt API.

Control playback and manage playlists via the Spotify API.

Fetch transcripts for YouTube videos.

Geocoding, directions, and places via Google Maps.

Chinese maps, geocoding, and routing via Amap.

Chinese maps, geocoding, and routing via Baidu.

Retrieve and triage Sentry issues.

Query and manage Stripe payments and customers.

Read and update HubSpot CRM records.

Query the Shopify Dev APIs and docs.

Search Airbnb listings and details.