Mcpbundles

What thinkchainai/mcpbundles MCP server does

The thinkchainai/mcpbundles MCP server provides the management interface for an MCP Bundles account. It lets an MCP-compatible assistant inspect, create, modify, and remove bundles; manage provider credentials; discover available providers; and review account usage. The Hub is intended for account and bundle administration rather than replacing the individual bundle endpoints used to run enabled provider tools.

A bundle groups tools from one or more providers. After a bundle is configured, MCP Bundles assigns it a URL such as https://mcp.mcpbundles.com/bundle/your-bundle-name. The Hub endpoint is https://mcp.mcpbundles.com/hub/.

How it works

The thinkchainai/mcpbundles MCP server authenticates requests against an MCP Bundles account. Authentication can use browser-based OAuth or a Bearer access token generated from the account dashboard. OAuth is described as the recommended option for Cursor, while an access token can be supplied for clients such as Claude Desktop.

A typical workflow is:

1. Connect an MCP client to the Hub endpoint.
2. Create a bundle and give it a name.
3. Add credentials for the providers the bundle should use.
4. Configure the bundle's provider tools.
5. Connect to the bundle-specific MCP URL to use those tools.

The Hub also provides provider search and provider-tool listing, so an assistant can inspect available integrations before configuring a bundle. Credentials are associated with the MCP Bundles account and are used by the configured bundle tools rather than exposed to the model.

Setup and configuration

An MCPBundles account and an MCP-compatible client are required. The README documents setup paths for Cursor and Claude Desktop, including a downloadable hub.mcpb package and a manual configuration that uses the Hub URL with an ACCESS_TOKEN environment variable.

Cursor can install the published .mcpb file from the GitHub release URL and then prompt for OAuth authentication. Claude Desktop can install the same file, or use the documented manual configuration. The repository also provides the Hub endpoint for clients that support remote MCP connections.

Do not commit access tokens to source control. The documented token is obtained from the MCP Bundles dashboard, and the service states that credentials use HTTPS/TLS, are stored in the account, and can be revoked.

Tools and capabilities

The thinkchainai/mcpbundles MCP server exposes tools with the mcpbundles_hub_ prefix.

• Bundle management: list, inspect, create, update, and delete bundles.
• Credential management: list, add, update, and remove provider credentials.
• Provider discovery: search providers, retrieve provider information, and list a provider's tools.
• Account monitoring: view account details and API usage statistics.

These tools support administration and discovery through the Hub. The README's examples show bundle-specific tools, such as GitHub, Slack, and Notion operations, being used after a bundle has been configured and its own MCP endpoint connected.

Limitations and notes

The Hub is authenticated and tied to an MCP Bundles account; it is not presented as an unauthenticated local server. OAuth and API-key-style Bearer authentication are both supported, but the setup method depends on the client.

The listed Hub tools do not themselves represent the complete catalog of provider actions. Provider actions become available through the separate MCP endpoint for each configured bundle. The repository describes more than 1,500 provider integrations, but the specific providers and tools available to an account depend on the bundles and credentials configured there.

The Hub also centralizes management, but it does not remove the need to configure provider credentials. A bundle may not be ready until its required credentials and tools have been set up.