If you only want the short answer: start with everything-claude-code for a broad Claude-oriented setup, use chrome-devtools-mcp when you need real browser debugging, and add playwright-mcpwhen you want repeatable automation instead of one-off inspection.
Updated: 2026-05-15 ¡ Query targets: best MCP servers, Claude Code MCP tools, open source MCP servers
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
Best for: operators who want a broad, Claude-oriented MCP setup fast
Strong fit for users who want a purpose-built Claude Code surface instead of stitching together many tiny tools first.
Chrome DevTools for coding agents
Best for: frontend debugging, DOM inspection, and runtime troubleshooting
If Claude Code needs to inspect the real browser, network requests, layout, or console state, this is the most direct path.
Playwright MCP server
Best for: repeatable end-to-end browsing, QA flows, and scripted task execution
Great when the goal is not just inspection but repeatable browser actions and workflows Claude can run reliably.
đ The fast, Pythonic way to build MCP servers and clients.
Best for: developers who want to expose internal tools or custom workflows to Claude Code
Useful when the highest-value server is your own. FastMCP lowers the cost of building and shipping that interface.
MCP Toolbox for Databases is an open source MCP server for databases.
Best for: teams that want a wider operational toolbox rather than one narrow integration
A practical pick when Claude Code needs access to many operational helpers instead of a single product-specific bridge.
| project | best use | category | signal |
|---|---|---|---|
| everything-claude-code | operators who want a broad, Claude-oriented MCP setup fast | MCP Servers | â163,083 |
| chrome-devtools-mcp | frontend debugging, DOM inspection, and runtime troubleshooting | MCP Servers | â36,585 |
| playwright-mcp | repeatable end-to-end browsing, QA flows, and scripted task execution | MCP Servers | â31,211 |
| fastmcp | developers who want to expose internal tools or custom workflows to Claude Code | MCP Servers | â24,733 |
| mcp-toolbox | teams that want a wider operational toolbox rather than one narrow integration | MCP Servers | â14,733 |
The highest-value MCP servers are usually not generic. They plug Claude Code into the exact surface you are working on: docs/context, browser/UI, database/cache, infra/runtime, observability/security, or research.
Context7 for up-to-date framework and library docs inside the coding loop.
Playwright for frontend flow testing, accessibility review, and browser automation.
Postgres for schema/index/query review and Redis for cache, TTL, hot-key, and rate-limit analysis.
Docker and Kubernetes for container, deployment, probe, autoscaling, and runtime inspection. Terraform for infra design review.
Grafana and Sentry for latency, alerts, and production errors. Semgrep and Trivy for application and image vulnerability review.
Firecrawl for docs extraction and architecture research, plus ArXiv for practical papers on distributed systems and backpressure topics.