MCP stands for Model Context Protocol. It is an open standard that lets AI assistants like Claude, ChatGPT, and others connect to external tools and data sources. Without MCP, an AI assistant can only think and write. With MCP, it can actually do things — search the web, generate images, look up live data, and interact with other services.
How does MCP work?
MCP works like a universal plug between your AI assistant and the tools it needs. Instead of each tool having its own setup process, MCP provides one standard way for tools to describe what they can do and for AI assistants to call them.
Here is what happens when you use an MCP tool:
- Your AI assistant connects to an MCP server (like ToolRouter).
- The server tells the assistant which tools are available and what each one does.
- When you ask the assistant to do something that needs a tool, it picks the right one and calls it.
- The tool runs and sends the result back to your conversation.
All of this happens inside your normal chat — you do not need to leave the conversation or open another app.
What is a connector?
In Claude, MCP connections are called connectors. When you add ToolRouter as a connector in Claude, you are setting up an MCP connection. The word "connector" is just Claude's name for it — the technology underneath is MCP.
Adding a connector takes about 30 seconds: go to Settings, then Connectors, then Add custom connector, and enter the ToolRouter URL. Once added, every ToolRouter tool becomes available in your Claude conversations on web, desktop, and mobile.
See Connect to Claude for step-by-step setup.
What about Claude Code and the terminal?
In Claude Code (the terminal version of Claude), MCP servers are added with a command instead of a settings menu. You run one command and ToolRouter becomes available to Claude Code automatically:
claude mcp add toolrouter -- npx -y toolrouter-mcpThe result is the same — Claude Code can discover and call any tool in the catalog. The only difference is how you set it up: a settings menu in Claude, a terminal command in Claude Code.
See Connect to Claude Code for the full guide.
Why does MCP matter?
Before MCP, giving an AI assistant access to a tool meant building a custom integration for each one. Every tool had its own authentication, its own data format, and its own setup process. If you wanted your AI to search the web, generate images, and check stock prices, you needed three separate integrations.
MCP replaces all of that with one standard. A single MCP connection to ToolRouter gives your AI access to hundreds of tools at once — image generation, video creation, web search, competitor research, lead finding, SEO analysis, and more. No separate signups, no separate API keys.
Which AI assistants support MCP?
MCP is supported by a growing number of AI assistants and development tools:
- Claude — via connectors (web, desktop, and mobile)
- Claude Code — via CLI command
- ChatGPT — via custom MCP apps (requires Developer mode)
- Codex CLI — via MCP config
- Microsoft Copilot — via Copilot Studio
- Cursor, Windsurf, VS Code, Cline — via MCP settings files
- Gemini CLI — via MCP config
- OpenClaw — via MCP config
How does ToolRouter use MCP?
ToolRouter is a hosted MCP server. Instead of running your own server or installing software, you point your AI assistant at ToolRouter's MCP endpoint and everything works immediately. ToolRouter handles authentication, billing, and tool discovery so you do not need to manage any of it yourself.
One connection gives your AI access to the full tool catalog. New tools are added every week and become available automatically — no updates or reinstalls needed.