Top 10 MCP Servers to Improve Your AI Workflows

In this guide, you will see:

• What an MCP server is and what this technology brings to the table.
• How to select the best MCP servers on the market.
• A list of the best 10 MCP servers to support your AI agents and workflows.

Let’s dive in!

What Is an MCP Server?

MCP, short for “Model Context Protocol,” is an open-source protocol that standardizes communication between LLMs and external tools/services. In other words, it is a universal language that AI agents can use to safely interact with any tool they need to complete a task.

An MCP server implements the MCP protocol, exposing a set of tools that an AI agent can discover and use. That is possible thanks to the SSE and Streamable HTTP technologies the protocol relies on.

Note that MCP is just one of many other AI protocols available.

The main benefits offered by integrating MCP servers into your AI workflows are:

• Standardization: They provide a consistent way for an AI application to interact with third-party tools. That eliminates the need to write custom integration code for each tool’s unique API, authentication, and data format.
• Discoverability: Agents can query an MCP server to understand what tools are available and how to use them. This enables dynamic and autonomous task execution, which stands at the core of every AI agent.
• Flexibility: Thanks to MCP servers, AI agents communicate only with the server, not directly with the end tools. This gives you control over which tools the AI can access.

How to Select the Best MCP Servers

The Awesome MCP Servers repository lists hundreds of useful MCP servers. With so many servers to choose from, selecting the best one for your needs requires clear evaluation criteria.

In particular, the key factors to keep in mind when selecting the best MCP servers are:

• Typical use cases: Is the server based on services you are interested in or already using? Does it provide tools that address your most common and time-consuming tasks? An MCP server’s value is measured by its ability to automate specific workflows.
• Key tools: The list of the most relevant tools exposed by the MCP server.
• Community trust and popularity: GitHub stars are a strong indicator of community adoption and trust. Also, a higher star count tends to correlate with a more stable and well-documented server. Other factors to consider are the number of contributors, recent contributions, and forks.
• License: MCP servers are generally open-source. In addition to the server’s own license, you should also review the licenses of any third-party software the MCP depends on. If you are already using those tools and they are covered under your existing licenses, you are good to go. Otherwise, you will need to assess their licensing terms and possibly allocate a budget for them.
• Programming language: The programming language used to develop the MCP server. This influences your requirements, installation, and eventual contributions to the source code.

The 10 Best MCP Servers

This section presents a curated list of the top MCP servers, selected and ranked according to the criteria outlined earlier.

1. Bright Data

Bright Data’s MCP server provides the data collection capabilities essential for any AI application. Since every AI system needs access to high-quality data, this MCP server is built to support virtually any use case.

Specifically, it equips AI agents and workflows with powerful tools to retrieve real-time web data. Those tools help AI ground its responses and interact accurately with web pages during task execution.

The great majority of AI-driven projects require fresh, up-to-date data to remain competitive and effective. That is why the Bright Data MCP server stands out as a top-tier solution.

Discover how to integrate it by following the official docs.

Typical use cases:

• Real-time information retrieval: You can ask for updated information. For example: “What are the top 5 trending news stories on The Guardian right now?”. The AI would use the search_engine tool with a query like "top stories The Guardian" to fetch the current headlines.
• Web scraping and data extraction: The server is designed for scraping tasks. It integrates with Bright Data’s Web Unlocker to access the content of any web page by bypassing all anti-bot measures. This means you can build an agent that performs web scraping tasks without worrying about being blocked.
• Bypassing geo-restrictions: Many websites display different content based on your geographical location or are accessible only from specific regions. To address that, the Bright Data MCP server can route its requests through a vast proxy network with over 150 million IPs distributed worldwide.
• Interactive browser automation: The server provides tools for “browser control,” which means an AI agent can do more than just download the raw HTML of a page. It can interact with a website dynamically.
• Structured data extraction: The server excels at retrieving structured data in JSON from major platforms thanks to the Web Scraper APIs. These can fetch real-time, structured data from Instagram, LinkedIn, Amazon, and many other sites via web scraping. For example, you can ask an agent to fetch the product details given an Amazon URL. The agent would then use a tool to retrieve that data.

Key tools:

• search_engine: Scrapes search results from Google, Bing, or Yandex. Returns SERP results in markdown.
• scrape_as_markdown: Scrapes a single webpage URL with advanced options for content extraction and gets back the results in Markdown. This tool can unlock any webpage, even if it uses bot detection or CAPTCHA.
• scrape_as_html: Scrapes a single webpage URL with advanced options for content extraction and gets back the results in HTML. This tool can unlock any webpage, even if it uses bot detection or CAPTCHA.
• session_stats: Tells the user about the tool usage during this session
• web_data_amazon_product: Quickly reads structured Amazon product data.
• web_data_amazon_product_reviews: Quickly reads structured Amazon product review data.
• web_data_linkedin_person_profile: Quickly reads structured LinkedIn people profile data.
• web_data_linkedin_company_profile: Quickly reads structured LinkedIn company profile data.
• web_data_zoominfo_company_profile: Quickly reads structured ZoomInfo company profile data.
• web_data_instagram_profiles: Quickly reads structured Instagram profile data.
• web_data_instagram_posts: Quickly reads structured Instagram post data.
• web_data_instagram_reels: Quickly reads structured Instagram reel data.
• web_data_instagram_comments: Quickly reads structured Instagram comments data.
• web_data_facebook_posts: Quickly reads structured Facebook post data.
• web_data_facebook_marketplace_listings: Quickly reads structured Facebook marketplace listing data.
• web_data_facebook_company_reviews: Quickly reads structured Facebook company reviews data.
• web_data_x_posts: Quickly reads structured X post data. Requires a valid X post URL.
• web_data_zillow_properties_listing: Quickly reads structured Zillow properties listing data.
• web_data_booking_hotel_listings: Quickly reads structured booking hotel listings data.
• web_data_youtube_videos: Quickly reads structured YouTube videos data.
• scraping_browser_navigate: Navigates a scraping browser session to a new URL.
• scraping_browser_go_back: Goes back to the previous page.
• scraping_browser_go_forward: Goes forward to the next page.
• scraping_browser_click: Clicks on an element.
• scraping_browser_links: Gets all links on the current page, text, and selectors.
• scraping_browser_type: Types text into an element.
• scraping_browser_wait_for: Waits for an element to be visible on the page
• scraping_browser_screenshot: Takes a screenshot of the current page
• scraping_browser_get_html: Gets the HTML content of the current page.
• scraping_browser_get_text: Gets the text content of the current page
• web_data_amazon_product_search: Quickly reads structured Amazon product search data.
• web_data_walmart_product: Quickly reads structured Walmart product data.
• web_data_walmart_seller: Quickly reads structured Walmart seller data.
• web_data_ebay_product: Quickly reads structured eBay product data.
• web_data_homedepot_products: Quickly reads structured homedepot product data.
• web_data_zara_products: Quickly reads structured Zara product data.
• web_data_etsy_products: Quickly reads structured Etsy product data.
• web_data_bestbuy_products: Quickly reads structured Best Buy product data.
• web_data_linkedin_job_listings: Quickly reads structured LinkedIn job listings data.
• web_data_linkedin_posts: Quickly reads structured LinkedIn posts data.
• web_data_linkedin_people_search: Quickly reads structured LinkedIn people search data.
• web_data_crunchbase_company: Quickly reads structured Crunchbase company data.
• web_data_facebook_events: Quickly reads structured Facebook events data.
• web_data_tiktok_profiles: Quickly reads structured TikTok profiles data.
• web_data_tiktok_posts: Quickly reads structured TikTok post data.
• web_data_tiktok_shop: Quickly reads structured TikTok shop data.
• web_data_tiktok_comments: Quickly reads structured TikTok comments data.
• web_data_google_maps_reviews: Quickly reads structured Google Maps reviews data.
• web_data_google_shopping: Quickly reads structured Google shopping data.
• web_data_google_play_store: Quickly reads structured Google Play Store data.
• web_data_apple_app_store: Quickly reads structured Apple App Store data.
• web_data_reuter_news: Quickly reads structured Reuters news data.
• web_data_github_repository_file: Quickly reads structured GitHub repository data.
• web_data_yahoo_finance_business: Quickly reads structured Yahoo Finance business data.
• web_data_youtube_profiles: Quickly reads structured YouTube profiles data.
• web_data_youtube_comments: Quickly reads structured YouTube comments data.
• web_data_reddit_posts: Quickly reads structured Reddit posts data.

To discover all tools, check out the Bright Data MCP server official repository.

Community trust and popularity: The repository has more than 700 stars. It is well-documented and widely adopted, as the 100+ forks demonstrate.

Compatible Websites: LinkedIn, real estate websites, Facebook, Reddit, YouTube, and many more.

Popular Integrations: N8N, Claude, Cursor, Perplexity, OpenAI, VS Code, Windsurf, and more.

License: The MCP Server is open-source (MIT license). Under the hood, it relies on Bright Data’s products, which come with a free trial.

Programming language: Node.js.

2. GitHub

GitHub MCP server is an indispensable tool for any team involved in software development. It helps AI agents become active participants in the development lifecycle. This means that they are capable of managing repositories, tracking issues, and even interacting with code.

Typical use cases:

• Automating GitHub workflows: Instead of manually clicking through GitHub’s interface, you can automate actions. For example, you could ask: “What’s the status of my latest PR?” and the bot could use the list_pull_requests and get_pull_request_status tools to find and report the answer.
• Extracting and analyzing data from GitHub repositories: This involves using the server to pull information from GitHub for analysis or reporting. For example, you could build a dashboard that lists all open pull requests and their current status.

Key tools:

• get_issue: Gets the contents of an issue within a repository.
• create_issue: Creates a new issue in a GitHub repository.
• add_issue_comment: Adds a comment to an issue.
• list_issues: Lists and filter repository issues.
• update_issue: Updates an existing issue in a GitHub repository.
• get_pull_request: Gets details of a specific pull request.
• list_pull_requests: Lists and filter repository pull requests.
• merge_pull_request: Merges a pull request.
• get_pull_request_diff: Gets the diff of a pull request.
• create_pull_request: Creates a new pull request.
• update_pull_request: Updates an existing pull request in a GitHub repository.
• delete_file: Deletes a file from a GitHub repository.
• list_branches: Lists branches in a GitHub repository.
• push_files: Pushes multiple files in a single commit.
• search_repositories: Searches for GitHub repositories.
• create_repository: Creates a new GitHub repository.
• fork_repository: Forks a repository.
• create_branch: Creates a new branch.
• run_workflow: Triggers a workflow via workflow_dispatch event.
• get_workflow_run: Gets details of a specific workflow run.
• get_workflow_run_logs: Downloads logs for a workflow run.
• rerun_workflow_run: Re-runs an entire workflow.
• rerun_failed_jobs: Re-runs only the failed jobs in a workflow run.
• cancel_workflow_run: Cancels a running workflow.

Discover all the available tools in GitHub’s official repository.

Community trust and popularity: The repository has more than 16.4k stars and is clearly documented. The high number of PRs (45+), contributors (60+), and forks (1.2k+) makes it a widely adopted and maintained repository.

License: This MCP server is open source (MIT license). It requires a GitHub account, which has free and paid tiers available.

Programming language: Go.

3. Supabase

Supabase is a popular open-source backend-as-a-service platform. Its MCP server gives AI agents full programmatic access to your project’s database, authentication, and storage. That opens the door to natural language-driven backend management.

Typical use cases:

• Database management and querying: You can interact with your Postgres database without writing SQL yourself. For example, you could ask, “How many users signed up last week?”. The agent would use the execute_sql` tool to run the necessary queries.
• Project and account administration: The server provides tools to manage your Supabase projects and organization settings. For instance, you can have an agent that creates a new project, pauses an inactive one, or gets a list of all your projects.
• Debugging and monitoring: When things go wrong, you can use the agent as a first line of defense. For example, you may ask, “Show me the API logs from the last hour to see if there are any errors”. To tackle that, the agent would use the get_logs tool.

Key tools:

• search_docs: Searches the Supabase documentation for up-to-date information.
• list_tables: Lists all tables within the specified schemas.
• list_migrations: Lists all migrations in the database.
• apply_migration: Applies a SQL migration to the database.
• execute_sql: Executes raw SQL in the database.
• get_project_url: Gets the API URL for a project.
• get_anon_key: Gets the anonymous API key for a project.
• generate_typescript_types: Generates TypeScript types based on the database schema.
• list_storage_buckets: Lists all storage buckets in a Supabase project.
• get_storage_config: Gets the storage config for a Supabase project.
• update_storage_config: Updates the storage config for a Supabase project (requires a paid plan).

The complete list of available tools is in the dedicated section of the repository.

Community trust and popularity: The repository has more than 1.7k stars. It is well-documented and widely adopted (150+ forks).

License: Open-source (Apache 2.0). Note that Supabase provides both free and paid plans.

Programming language: Node.js.

4. Playwright

The Playwright MCP server enables LLMs to interact with web pages by taking advantage of the Playwright browser automation API. Behind the scenes, it relies on accessibility snapshots. That overcomes the need for screenshots and AI models with visual capabilities.

Learn more about what Playwright has to offer.

Typical use cases:

• Human-like interaction: Your AI agent will gain the ability to interact with web pages, such as clicking, navigating, taking screenshots, and more. This means you can build advanced AI agents that perform real-world tasks on websites using just a prompt (e.g., similar to what you can achieve with Browser-Use).
• Automated test generation: To do so, you have to describe a complete user journey. The AI can use its browser control tools to perform the steps and use the browser_generate_playwright_test tool to output a Playwright test script.

Key tools:

• browser_snapshot: Captures accessibility snapshot of the current page.
• browser_click: Performs a click on a web page.
• browser_drag: Performs drag and drop between two elements.
• browser_hover: Hovers over the element on the page.
• browser_type: Typse text into an editable element.
• browser_select_option: Selects an option in a dropdown.
• browser_wait_for: Waits for text to appear or disappear or for a specified time to pass.
• browser_navigate: Navigates to a URL.
• browser_pdf_save: Saves page as PDF.
• browser_tab_list: Lists browser tabs.
• browser_tab_new: Opens a new tab.
• browser_tab_select: Selects a tab by index.
• browser_tab_close: Closes a tab.
• browser_generate_playwright_test: Generates a Playwright test for the given scenario.
• browser_screen_move_mouse: Moves the mouse to a given position.
• browser_screen_click: Clicks the left mouse button.
• browser_screen_drag: Drags the left mouse button.
• browser_screen_type: Types text.
• browser_press_key: Presses a key on the keyboard.

The tools section in the Playwright repository describes them all.

Community trust and popularity: The 13.1k+ GitHub stars demonstrate that this repository is widely trusted and adopted. It also has several contributors and forks (900+), as well as good documentation.

License: Open-source (Apache 2.0).

Programming language: Node.js.

5. Notion

Notion’s MCP server turns Notion into a dynamic knowledge base that AI agents can read from and write to. This gives your AI the automation power to handle tasks like documentation, project management, and content creation.

Typical use cases:

• Automated task management: You can manage your project plans using natural language. The AI will create a new entry in your database with the properties you prompted.
• New Notion database creation: The AI agent can create new Notion databases. You could use a prompt like: “Create a new database to track my job applications. It should have columns for the company name, position, and a link to the job description”.
• Knowledge retrieval: You can find solutions to technical problems by asking the agent to search across all the documentation.

Key tools:

• Search: Finds anything in your Notion workspace, connected apps, or the web by asking questions in plain English.
• Search by Title: Fallback search tool when AI subscription isn’t available. Performs keyword search on page titles only.
• View: Looks at any page, database, file, or user in your Notion workspace to see what is inside.
• Get Comments: Lists all comments on a specific page or block, including threaded discussions.
• Get User: Gets detailed information about a specific user by their ID or reference.
• Create Pages: Makes new pages in your workspace with any content you want. Specify where you would like this page to be added, or it will be a default private page.
• Create a comment: Adds a comment to a page or block.
• Update Page: Edits existing pages by changing their title, content, or other properties.

The Notion MCP server tools documentation describes when to use each tool and provides useful prompts to get you started immediately.

Community trust and popularity: It has 2.3k+ stars and good documentation. It also has a good amount of recent contributions, making it well-maintained and adopted.

License: Open-source (MIT license). Notion itself provides a wide array of features with the free plan, but some are provided for paying users.

Programming language: Node.js.

6. Atlassian

Atlassian MCP server is specifically created for automating workflows that involve Confluence, Jira, Jira Cloud, and Server/Data Center deployments using LLMs. Those solutions are widely used for documentation, issue tracking, and team collaboration. Thus, integrating them with AI enables intelligent agents to manage tickets, update documentation, and more.

Typical use cases:

• Smart Jira automation: By properly prompting the LLM, you can automate Jira workflows for information retrieval and search, and issue creation and management.
• Confluence documentation management: You can turn Confluence into a dynamic knowledge base you can talk to. You can create and manage documentation directly from your chat. You can also ask the agent to search and summarize documentation for you.

Key tools:

• jira_search: Searches Jira issues.
• jira_create_issue: Creates a new Jira issue.
• jira_update_issue: Updates an existing Jira issue.
• confluence_search: Searches Confluence content.
• confluence_get_page: Gets the content of a specific page.
• confluence_create_page: Creates a new page.
• confluence_update_page: Updates an existing page.

Read all the available tools in the Atlassian MCP server repository section.

Community trust and popularity: It has 2.1k+ stars and good documentation. It also has lots of contributors (50+) and forks (360+), making it well-maintained and adopted.

License: Open-source (MIT license). Note that Jira and Atlassian come with both free and paid plans.

Programming language: Python, but distributed only via Docker.

7. Serena

Serena MCP server is a coding agent toolkit that works directly on your codebase. It provides all the tools that are typical of an IDE’s capabilities. In particular, it:

• Uses LSP (Language Server Protocols) to parse and understand code semantically.
• Can read and write code, but also execute shell commands.
• Has a persistent understanding of a specific codebase, as it features an onboarding and memory system.

Typical use cases:

• LLM-powered coding: Serena is designed for any coding task. It can read, write, and execute code. It can also read logs and the terminal output. It has direct and indirect support for programming languages like Python, JavaScript, Go, and more.

Key tools:

• activate_project: Activates a project by name.
• create_text_file: Creates/overwrites a file in the project directory.
• delete_lines: Deletes a range of lines within a file.
• delete_memory: Deletes a memory from Serena’s project-specific memory store.
• execute_shell_command: Executes a shell command.
• find_symbol: Performs a global (or local) search for symbols with/containing a given name/substring (optionally filtered by type).
• get_active_project: Gets the name of the currently active project (if any) and lists existing projects
• get_current_config: Prints the current configuration of the agent, including the active modes, tools, and context.
• get_symbols_overview: Gets an overview of the top-level symbols defined in a given file or directory.
• initial_instructions: Gets the initial instructions for the current project.
• insert_after_symbol: Inserts content after the end of the definition of a given symbol.
• insert_at_line: Inserts content at a given line in a file.
• list_dir: Lists files and directories in the given directory (optionally with recursion).
• list_memories: Lists memories in Serena’s project-specific memory store.
• prepare_for_new_conversation: Provides instructions for preparing for a new conversation.
• read_file: Reads a file within the project directory.
• read_memory: Reads the memory with the given name from Serena’s project-specific memory store.
• replace_lines: Replaces a range of lines within a file with new content.
• search_for_pattern: Performs a search for a pattern in the project.
• summarize_changes: Provides instructions for summarizing the changes made to the codebase.

Discover the full list of tools in Serena’s repository.

Community trust and popularity: Serena has 2.9k+ stars and a very wide documentation.

License: Open-source (MIT license).

Programming language: Python.

8. Filesystem

The filesystem MCP server repository is designed to create AI agents that manage your filesystem operations.

Key tools:

• read_file: Reads the complete content of a file.
• write_file: Creates a new file or overwrites an existing one.
• create_directory: Creates a new directory or ensures it exists.
• move_file: Moves or renames files and directories.

Access the comprehensive list of available tools in the repository.

Community trust and popularity: This repository has 56k+ stars, wide documentation, and a high number of contributors (580+). Note that these metrics are related to the repository of the entire project, which lists several MCP servers.

License: Open-source (MIT license).

Programming language: The repository provides servers in Node.js and Python.

9. Figma

The Figma MCP server is designed to give Cursor access to your Figma files, using AI-powered coding tools. Its goal is to shorten the time you need to create one-shot designs, without pasting screenshots.

Typical use cases:

• Decrease POC time: You would use this server to decrease the time needed to create a POC (Proof of Concept) of a Figma design by prompting an LLM and leaving the AI agent to do the job autonomously.

Key tools:

• get_code: Provides a structured React + Tailwind representation of your Figma selection.
• get_variable_defs: Extracts the variables and styles used in your selection.

Community trust and popularity: The repository has 8.6k+ stars.

License: Open-source (MIT). Instead, Figma provides free and paid plans.

Programming language: Node.js.

10. Grafana

Grafana is an open-source platform for data visualization, monitoring, and analysis. It is particularly appreciated for its agnosticism, extendibility, and because it is an open system. This means that it acts as a central hub that connects to dozens of different data sources. The Grafana MCP server allows you to create an AI agent that interacts with the whole Grafana ecosystem.

Typical use cases:

• AI incident management: You can create agents to help you manage the entire incident lifecycle directly from your chat interface. It also lowers the entry barrier for querying systems by translating natural language into specific PromQL or LogQL queries.

Key tools:

• search_dashboards: Searches for dashboards.
• query_prometheus: Executes a query against a Prometheus datasource.
• list_incidents: Lists incidents in Grafana Incident.
• query_loki_logs: Queries and retrieves logs using LogQL.
• get_analysis: Retrieves a specific analysis from a Sift investigation.

Automate your Grafana incident management using all the available tools.

Community trust and popularity: 1k+ stars, good documentation, and frequent contributions.

License: Open-source (Apache 2.0). For Grafana, consult their pricing page.

Programming language: Go.

Best MCP Servers: Summary Table

Below is a summary table for a wide overview of the MCP servers you discovered in this article:

Now that you know the best MCP servers, you can learn how to use them by reading these guides:

• Integrate Qwen-Agent with MCP to Build Agents with Live Data Access
• Web Scraping With MCP Servers: A Step-by-Step Guide
• Integrating Google ADK with an MCP Server for AI Agent Development
• Building Web Scraping Agents with CrewAI & Bright Data’s Model Context Protocol (MCP)

Conclusion

In this article, you learned what an MCP server is and why it is useful. You explored the main factors to consider when choosing the best MCP servers on the market. Then, you saw how those criteria apply in a curated list of the top 10 available options.

Among the options listed, Bright Data’s MCP server stands out as one of the best. The reason is simple: every AI project or workflow depends on high-quality data!

That is exactly where the Bright Data MCP server excels. It gives AI the ability to ethically retrieve the data it needs from the Web, the largest and richest source of information on the planet.

Now, when building production-ready AI workflows, you need tools that can retrieve, validate, and transform web content reliably. That is precisely what you can find in Bright Data’s AI infrastructure.

Create a Bright Data account and try all our products and services for AI development!