Installs into .claude/skills of the current project.
Are you the author of Build Authenticated Mcp Tools And Servers With Arcade Mcp?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/agentskillexchange-build-authenticated-mcp-tools-and-servers-with-arc)
---
name: "Build authenticated MCP tools and servers with Arcade MCP"
slug: "build-authenticated-mcp-tools-and-servers-with-arcade-mcp"
description: "Use Arcade MCP to create custom MCP servers and tools with OAuth-aware authorization, evals, and deployment paths for agent tool-calling workflows."
github_stars: 895
verification: "security_reviewed"
source: "https://github.com/ArcadeAI/arcade-mcp"
author: "ArcadeAI"
publisher_type: "company"
category: "Integrations & Connectors"
framework: "MCP"
tool_ecosystem:
github_repo: "ArcadeAI/arcade-mcp"
github_stars: 895
---
# Build authenticated MCP tools and servers with Arcade MCP
Use Arcade MCP to create custom MCP servers and tools with OAuth-aware authorization, evals, and deployment paths for agent tool-calling workflows.
## Prerequisites
Python, arcade-mcp, an MCP-compatible client, optional Arcade Cloud account for hosted auth and deployment
## Installation
Use the upstream install or setup path that matches your environment:
- uv tool install arcade-mcp
- uv run server.py # stdio (default), for Claude Desktop and CLI tools
- uv run server.py http # HTTP+SSE, for Cursor and VS Code; docs at http://127.0.0.1:8000/docs
- git clone https://github.com/ArcadeAI/arcade-mcp.git
Requirements and caveats from upstream:
- [](https://pypi.org/project/arcade-mcp/)
- **Open-source Python framework for building MCP servers and tools.**
- arcade-mcp is the Python framework for building [Model Context Protocol](https://modelcontextprotocol.io) servers and the tools that run inside them. It powers the 7,500+ prebuilt tools across 81 MCP servers at [Arcad...
Basic usage or getting-started notes:
- The secrets (OAuth tokens, API keys, etc) are securely injected by Arcade into your tool call at runtime, so that your tool can authorize requests to upstream APIs, for example. **The client and the LLM never see the...
- bash
- ### Scaffold a new server
- Source: https://github.com/ArcadeAI/arcade-mcp
- Extracted from upstream docs: https://raw.githubusercontent.com/ArcadeAI/arcade-mcp/HEAD/README.md
## Documentation
- https://docs.arcade.dev
## Source
- [Agent Skill Exchange](https://agentskillexchange.com/skills/build-authenticated-mcp-tools-and-servers-with-arcade-mcp/)