📄️ How MCPBundles Works
Understanding the architecture from MCP servers to tools to providers. Learn how the four-layer system connects your AI assistants to real-world APIs.
📄️ What Are A2A Agents?
Learn how MCPBundles A2A agents work — the Agent-to-Agent protocol, Agent Cards, scheduled runs, webhooks, and how they differ from Studio chat.
📄️ MCP Servers
What MCP servers are on MCPBundles, how they're organized, and their lifecycle. Learn about categories, publication status, readiness states, and how to manage them.
📄️ The Hub Endpoint
One MCP URL for all your tools. The Hub aggregates every enabled MCP server into a single endpoint with dynamic discovery and execution.
📄️ Execution Modes (Direct vs Dynamic)
Understand how MCP servers run in Direct vs Dynamic mode, why Dynamic exists, and the 100-tool limit for Direct.
📄️ Understanding Providers
External APIs, authentication methods, and credential management. Learn how to connect providers, manage OAuth tokens, API keys, and verify credentials.
📄️ Understanding Workspaces
How workspaces organize your MCPBundles environment. Learn about personal vs team workspaces, what's shared, permissions, and how the workspace boundary provides privacy.
📄️ MCP Resources & File Storage
Understanding MCP Resources and how to use file storage with your AI
📄️ MCPB Files (.mcpb): Format Reference, Manifest, and Examples
MCPB (.mcpb) is an open packaging format — a ZIP archive plus manifest.json that lets MCP clients like Claude Desktop install a local MCP server in one click. Full structure, fields, and examples.
📄️ Understanding Skills
What SKILL.md files are, how they work with MCP servers, and why domain knowledge makes AI assistants more effective at using tools.