MCP Client
An MCP client that brings external AI tools into your workspace
Connect external MCP servers to your Runwork workspace and your agents gain their capabilities. Connect a weather service, a code analysis tool, a knowledge base, anything with an MCP server. External tools appear alongside your apps and can be loaded as skills in any connected agent.
docs.internal.acme.com/mcp code-review.tools.dev/mcp weather.api.tools/mcp 
Powered by Runwork AI What MCP Client gives you
Connect Any MCP Server
Add external MCP server URLs in workspace settings. Your agents immediately discover and can use their tools. No code, no configuration beyond the URL.
Auto-Discovery
When you connect an MCP server, Runwork automatically discovers available tools, resources, and prompts. They appear in your workspace ready to use.
Unified Tool Set
External MCP tools work alongside your native workspace tools. Your agent can query your CRM (native) and check the weather (external MCP) in the same conversation.
External Tools as Skills
Connected MCP tools automatically generate skill descriptions. Browse them with slash commands in your agent, and use them like any other skill.
Domain Allowlisting
Security first. Approve which external domains your AI can call. Manage the allowlist from workspace settings. Rate limiting built in.
HTTP Request Tool
Skills that reference external APIs get a secure HTTP request tool. The AI can call approved endpoints directly, with full request and response handling.
Access from anywhere
Desktop & Web
Codex DEFAULT 
Claude Code 
Cursor Browse connected MCP servers and their available tools. Test tool calls directly from the dashboard.
CLI
Why MCP Client matters
- Extend your workspace with any MCP-compatible service
- External tools feel native: same slash commands, same skill system
- Secure by default with domain allowlisting and rate limiting
- The AI ecosystem flows both ways: out via MCP Servers, in via MCP Client
How MCP Client works
The MCP Client is the other half of Runwork's AI ecosystem. While MCP Servers let external tools access your workspace, the MCP Client lets you bring external tools in. Connect any MCP server URL and your agents gain new capabilities instantly.
Adding an external MCP server is simple: paste the server URL in your workspace settings and Runwork connects, authenticates, and discovers all available tools, resources, and prompts. These show up alongside your native workspace tools. Your agent can use them in any conversation without you having to switch contexts or tools.
External MCP tools automatically become skills. This means you can browse them with slash commands and load them into conversations just like your vibe-app skills. Your agents can orchestrate across native apps and external services in the same conversation. Connected MCP servers are available across all surfaces. Your local AI agents access them automatically after runwork sync. The web dashboard lets you test tool calls directly.
Security is built in. Every external domain must be explicitly approved in your workspace allowlist before your agents can make calls. Rate limiting prevents abuse. You manage the allowlist from workspace settings, and it applies to both MCP connections and direct HTTP requests from skills.
The MCP Client also enables a powerful pattern: connect your company's internal MCP servers to centralize knowledge. A documentation server, a metrics server, an HR system. Your Runwork workspace becomes the single point your agents can access and orchestrate across all of them, combined with your vibe-coded apps, integrations, and workflows.