Skip to directory

About MCPBundles Hub

MCPBundles Hub is a hosted meta-MCP that discovers connected MCP servers and dynamically exposes their tools through one endpoint, with sandboxed code execution for orchestration.

What you can use it for

  • Discover servers and list connected servers.
  • Inspect a tool schema and test server credentials.
  • Execute multi-server workflows through sandboxed Python.
Connection and setup
Hosting
remote
Transport
Streamable HTTP
Endpoint
https://mcp.mcpbundles.com/hub/
Authentication
OAuth dashboard authentication; bearer token setup is available for compatible CLI/client workflows.

Requirements

  • MCPBundles account and workspace.
  • MCP-compatible client with remote Streamable HTTP support, or Python tooling for the CLI path.
  • Enable integrations and connect each provider's OAuth/API-key credentials in the dashboard before using its tools.

Setup steps

  1. Create/sign in to a MCPBundles account.
  2. Enable the desired catalog servers and configure their required provider credentials in the dashboard.
  3. Add https://mcp.mcpbundles.com/hub/ to a supported client; complete OAuth when prompted. For clients requiring a local bridge, use the official repository's mcp-remote Streamable HTTP configuration and supply a dashboard access token through an environment placeholder.
  4. Alternatively install the CLI with pip install mcpbundles and run mcpbundles init; authenticate when prompted.
  5. Use the Hub discovery/catalog tools to find enabled servers, then discover exact service tool names and schemas before calling them. Test with Hub status/readiness checks.
Tools 6

The Hub itself exposes six exact meta-tools; it does not statically expose all connected-server tools. Its dynamic sandbox can list and invoke enabled server tools with server-specific credential bindings, so actual available tools depend on workspace connections and permissions.

  • list_connected_mcp_servers

    List MCP servers connected to the current workspace.

    01
  • discover_mcp_servers

    Discover MCP servers available to connect from the directory.

    02
  • describe_tool

    Retrieve the complete schema for a named tool.

    03
  • code_execution

    Execute Python in the Hub sandbox to discover and call tools from connected servers.

    04
  • test_connected_mcp_server

    Test whether a connected MCP server’s credential binding works.

    05
  • health_check

    Diagnose Hub connectivity and health.

    06
Limitations and notes
  • code_execution runs user-authenticated connected server tools in a Python sandbox (120-second timeout); connected actions retain each upstream server’s permissions and side effects.
Source discrepancies
  • Discovery queue source_urls incorrectly pointed at https://github.com/github. The correct publisher repository found is https://github.com/thinkchainai/mcpbundles.
  • The Hub endpoint is separate from MCPBundles bundle-specific endpoints; those are not merged into this record.
  • Official web docs recommend built-in OAuth; the release repository also documents bearer-token setup through mcp-remote. These are alternate documented client flows, not conflicting endpoint identities.
Original sources

Documentation checked 2026-10-28T00:00:00Z. Connection has not been runtime tested.

Similar tools