Discover MCP servers, skills, and plugins that give
your AI agents the power to do more.
Last updated: October 17, 2026 · 12 components · See the changelog
The right tools. A stronger agent. A better starting point.
Get to know MCP, skills, and plugins. No experience needed.
Turn any website into clean, structured markdown. Built by Mendable for crawling JavaScript-heavy pages and bringing the live web into Claude, Cursor, or Windsurf. The most useful MCP server for research and content extraction; paid plans scale with how much you crawl.
Bring your repositories into the agent. Search code, triage issues, open pull requests, and assist with code review from inside Claude, Cursor, or any MCP client. Free and open source; permissions follow whatever GitHub token you provide, so use a fine-grained PAT scoped to what the agent should see.
Query document databases from your agent. Browse collections, inspect schemas, and pull data into grounded workflows. Free; bring your own connection string and trust whatever schema you already have - the agent sees what your database sees.
Connect your agent to AWS infrastructure workflows. Read resource state, pull operational metrics, and chain actions across services. Free and open source; treat any agent-driven AWS call as production with the corresponding blast radius - identity is everything here.
Work with containers and images from your agent. Inspect logs, manage dev environments, and run side-effecting operations from inside an MCP client. Free; container access inherits the host permissions for whatever the Docker socket exposes, so be deliberate about what you mount.
Give your agent an isolated code-execution environment. Run Python in a sandbox, do data analysis, and test scripts without trusting local state. Freemium; great for untrusted code, but every execution is remote, metered, and rate-limited.
An MCP-flavored suite for SEO audits, SERP scraping, web summarization, company intelligence, and email health checks. Built by the SiteTrace team. Freemium with an API key; a fast on-ramp if you do not want to assemble those capabilities yourself.
Connect your agent to global payments. Inspect transactions, manage payment methods, and handle disputes through Adyen. Freemium and enterprise-tier; treat any agent-driven payment operation as production, and put human review on anything that moves money.
Query records, create and update rows, manage views, and run automations from your agent. Beginner-friendly and a good first MCP server. Freemium; bring your own API key and consider scoping tokens to specific bases rather than the whole workspace.
Search Algolia indexes with ranked results, autocomplete, and typo tolerance. Useful if you already use Algolia for search; less useful if you do not. Freemium; the agent can read and write your search configuration, so be careful with the API key.
Add flashcards, search your deck, generate cards from text, and trigger review sessions from your agent. Free and open source; a great way to turn agent output into long-term memory via spaced repetition, especially for studying.
Access Claude models, prompt caching, batch completions, and usage inspection from your agent. Free at the MCP layer; the cost is whatever Anthropic charges for the underlying tokens. The most direct way to make a non-Claude agent call Claude as a sub-task.
Documentation, maintenance, and security signals - not just star counts.
Understand permissions, dependencies, and the tool's intended capabilities.
Real context for experienced builders and people taking their first step.