MCP servers and AI agent tools worth installing
Let Claude, Codex and Cursor read files, browse and drive a browser.
Allow AI agents to read internet content, supporting platforms like Twitter, Reddit, Bilibili
No API fees, automatically handles login and blocks
InstallHelp me install Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.mdReal-time global intelligence dashboard, AI news aggregation and geopolitical monitoring
Integrates multiple data sources, no API key required
AI job search agent that scans job boards and optimizes application materials
Runs locally, no personal data uploaded
A set of skills and a development workflow for Claude Code and other agents: brainstorm, write a plan, TDD, then dispatch subagents to implement and review.
Makes "plan before coding" a process agents follow automatically. One of the most installed skill packs of 2026.
InstallIn Claude Code, run /plugin marketplace add obra/superpowers-marketplace, then /plugin install superpowers@superpowers-marketplaceAn MCP server that lets Claude Code, Cursor and similar tools fetch up-to-date official docs and examples for libraries, so AI writes fewer outdated APIs.
Set it up once, then add "use context7" to a prompt. Especially useful for fast-moving frameworks like Next.js and React.
Google's official MCP server that lets coding agents control a real Chrome: read the console, capture network requests, run performance profiles, take screenshots.
After a frontend change the agent can open the browser to verify and debug on its own, with DevTools-level detail beyond screenshot-only browser tools.
A local proxy that lets Claude Code and other agents use DeepSeek, Kimi, GLM, Gemini, OpenRouter or local models, routing automatically by task.
Maintained by a Chinese developer; especially useful for users who can't access Anthropic directly or want lower costs. A desktop version is available.
An MCP toolkit built on language servers that lets agents read and edit code by finding symbols, references and functions instead of reading whole files.
Saves tokens and reduces bad edits in large projects. Works with Claude Code, Codex, Cursor and JetBrains.
Reads local Claude Code, Codex and similar logs and reports token usage and estimated cost by day, month and session.
Runs with one command, no install. Lets subscribers see their usage and 5-hour limit consumption. Now rewritten in Rust.
Updated Oct 25, 2026 · ★ GitHub