MCPBundles
- ide
Studio is where you run MCP tools in the browser against real credentials -- see inputs, outputs, and errors before you trust an agent. Enable bundles in your workspace, then add your Hub or per-bundle streamable HTTP URL in Cursor, Claude, ChatGPT, or VS Code. Workspace-scoped c
About
What is MCPBundles?
MCPBundles is a platform for using MCP (Model Context Protocol) servers. It provides an in-browser Studio for testing tool execution, inspecting schemas, and debugging failures before connecting to external hosts via streamable HTTP MCP URLs. It also offers a CLI for terminal discovery and scripting. MCPBundles is for developers and teams who want to validate and deploy MCP servers into AI assistants, IDEs, and chat applications.
How to use MCPBundles?
1. Enable the bundles you need and connect accounts (see getting started guide).
2. Open Studio from a bundle to run tools with live data.
3. When satisfied, add your MCP URL in the host you use every day (see integration guides for Cursor, Claude Desktop, ChatGPT, VS Code, Windsurf).
4. Optionally use execution modes (Hub vs bundle) to control surface area exposed to the model.
5. Optionally use the mcpbundles CLI for terminal discovery and scripting.
Key features of MCPBundles
- Interactive Studio for testing tools and debugging failures
- Validate credentials and confirm provider wiring
- Streamable HTTP MCP URLs for connecting to external hosts
- Integration guides for Cursor, Claude, ChatGPT, VS Code, Windsurf
- MCP Apps: dashboards, charts, and interactive UI within AI conversations
- CLI for terminal discovery and scripting
Use cases of MCPBundles
- Test MCP server tools with live credentials before any agent uses them
- Debug tool schemas and responses in an isolated browser environment
- Deploy MCP servers to multiple AI hosts using a single bundle URL
- Build and share interactive MCP Apps (charts, dashboards) inside AI chats
FAQ from MCPBundles
Is MCPBundles only a dashboard?
No — the differentiator is Studio, which lets you execute and debug tools before agents rely on them.
Is MCPBundles the same app as Cursor or Claude?
No. You configure access and test tools in MCPBundles, then paste your Hub or bundle URL into each host (see integration guides above).
How do I validate credentials?
See the credential validation guide and run tools in Studio to confirm they work with live data.
Details
- Author
- thinkchainai
- Category
- ide
- Repository
- thinkchainai/mcpbundles
Overview
Studio is MCPBundles' in-browser environment for using your MCP servers the way an agent will: execute tools with real credentials, inspect schemas and responses, and debug failures -- before you wire the Hub or a per-bundle URL into an external host.
What Is Studio?
Studio is an interactive testing surface for bundles you've enabled in a workspace. It is the fastest way to learn what each tool accepts and returns, validate credentials, and confirm providers are wired correctly. Full guide: Studio.
How Do You Use MCP Servers With MCPBundles?
1. Get started -- enable the bundles you need and connect accounts.
2. Open Studio from a bundle (see Studio guide) and run tools with live data.
3. When you are satisfied, connect your AI by adding your MCP URL in the host you use every day.
4. Optional: use execution modes (Hub vs bundle) to match how much surface area you expose to the model.
5. Optional: mcpbundles CLI on PyPI for terminal discovery and scripting.
Where Do You Connect MCPBundles In Your IDE Or Chat App?
Add your streamable HTTP MCP URL in the host's MCP settings. Step-by-step guides:
- Cursor
- Claude Desktop / Claude Code
- ChatGPT
- VS Code
- Windsurf
MCP Apps: Charts And Interactive UI In The Host
When the host supports it, tools can return MCP Apps -- dashboards, charts, and workflows inside the conversation. Read the launch post and the explainer MCP Apps -- interactive UI inside AI.
More On The Product
- Browse skills / bundles -- discovery paths into the catalog.
- Use cases -- how teams deploy MCP with MCPBundles.
- What is MCP? -- protocol primer, including MCP Apps.
- FAQ / Troubleshooting connectivity
FAQ
- Is MCPBundles only a dashboard? No -- the differentiator for using servers day-to-day is Studio for tool execution and debugging before agents rely on them.
- Is MCPBundles the same app as Cursor or Claude? No. You configure access and test tools in MCPBundles; you still paste your Hub or bundle URL into each host (see integration links above).
- How do I validate credentials? See credential validation and run tools in Studio.
