What are the best MCP servers to connect to an AI agent?
Start with servers for the systems you already query by hand: your database, your error tracker, your issue tracker, your deploys. MCP is the standard that lets any compatible assistant call them without custom glue.
Surveyed 17 September 2026
Connect tools with MCP
Open in Tools →- 01Chrome DevTools MCP
MCP server that exposes Chrome DevTools so agents can inspect, debug, and profile web pages programmatically.
AI Agents - 02Playwright MCP
MCP server from Microsoft that lets agents drive a browser through Playwright tool calls.
AI Agents - 03Context7
Serves current, version-specific library docs and code examples to coding agents over MCP, so models avoid stale APIs.
Developer Tools - 04GitHub MCP Server
Official MCP server connecting coding agents to GitHub repositories, issues, pull requests, and Actions.
Developer Tools - 05Figma MCP Server
Connects Figma files to AI coding agents over MCP, exposing design context and writing content back to the canvas.
Developer Tools - 06Notion MCP Server
Official MCP server for the Notion API, letting AI agents read, create, and edit pages and databases.
Developer Tools - 07Slack MCP Server
MCP server that reads Slack channels and DMs and posts messages without requiring admin permissions.
Developer Tools - 08Supabase MCP
Server that lets AI agents query, mutate, and manage Supabase projects through the Model Context Protocol.
Developer Tools - 09Sentry MCP
MCP server that lets Claude and other agents read issues, traces, and replays from a Sentry organization.
AI Agents - 10Stripe MCP
Official MCP server letting agents search Stripe docs, inspect account data, and call the API with scoped keys.
Developer Tools - 11Cloudflare MCP Server
Official MCP servers for managing Workers, KV, R2, D1, and DNS, with analytics, logs, and docs search from any client.
Developer Tools - 12Vercel MCP
Hosted MCP server giving agents access to Vercel projects, deployments, and build logs to investigate failed builds.
Developer Tools
A curated selection in editorial order. Use the fit and evidence to judge it for your task. Something missing?
What to look for
- 01Are the permissions scoped and read-only where possible? An MCP server inherits whatever credentials you give it.
- 02How many tools does it expose? Servers with dozens of tools crowd the context window and degrade tool choice.
- 03Does the tool output stay small? Servers that return unbounded payloads exhaust the window in one call.
Common questions
- What is MCP?
- The Model Context Protocol, an open standard for exposing data and actions to AI assistants. Write one server and any compatible client can use it, instead of building a separate integration per assistant.
- What are the security risks of MCP servers?
- The server runs with the credentials you hand it, and data it returns can carry injected instructions. Scope permissions tightly, prefer read-only, and treat everything a server returns as untrusted content.
More in Work with agents
- Code with an agentWork across a repository from an issue, prompt, or terminal session.
- Control a browserLet an agent navigate, inspect, test, and complete workflows on the web.
- Research with an agentFind sources, verify claims, synthesize evidence, and preserve citations.
- Give an agent memoryPersist project context, decisions, preferences, and reusable knowledge.
- Find agent skillsMarketplaces and indices for discovering, vetting, and installing agent skills.