Join waitlist
Back to products
Secure MCP Framework by Arcade.dev

Secure MCP Framework by Arcade.dev

Skip migration and launch MCP with built-in Auth

Website arcade.dev
Overview

What it is

arcade-mcp is the MCP framework that grows with you. Launch a secure MCP server with built in Auth, zero setup, and no rewrites. Write clean Python, scale to production, and connect tools like Gmail or Slack in one line. Open source, eval ready, and works with LangGraph, Cursor, Claude, and more.

Intent

I need it when

Scale MCP server infrastructure from development to production without code rewrites

MIT-licensed open-source framework runs locally offline-friendly during development, then scales to Arcade-hosted production deployment with managed secrets and audit logs, maintaining code compatibility throughout

Integrate AI agents with existing IDEs and LLM platforms (Cursor, Claude, LangGraph)

Framework is agent-compatible with major platforms and provides arcade configure command for seamless MCP server connection to developer tools, enabling agents to access custom tools across preferred development environments

Securely authenticate users and manage credentials for AI agent integrations

Framework provides built-in OAuth support, secrets management via .env locally and managed secrets in production, and enterprise-grade authentication (SSO/SAML on Enterprise plan) so agents can safely access external services like Google, Slack, and Salesforce

Build custom MCP servers quickly with minimal setup overhead

Arcade MCP Framework enables developers to spin up production-ready MCP servers with one command using clean Python code, eliminating boilerplate and reducing time-to-deployment from weeks to minutes

Deploy AI agents that can take real actions on behalf of users across multiple platforms

Arcade's runtime layer enforces agent authorization and enables authenticated actions in third-party services, transforming AI assistants from passive information tools into agents that can execute transactions and workflows

Drop

Not a fit when

  • User needs a fully managed, no-code AI agent platform without framework-level customization
  • Organization requires on-premises deployment only and cannot use Arcade-hosted infrastructure
  • Team lacks Python development expertise and cannot write custom MCP server code
  • Use case requires real-time agent actions without OAuth or authentication workflows
  • Budget is extremely constrained and cannot accommodate any usage-based charges beyond free tier limits
Commercials

Pricing

Usage-based with tiered plans: Free tier with included allocations, Growth plan at $25/month plus per-use charges, Enterprise with custom pricing View pricing