In the podcast, Toby Padilla discusses the development and growth of GitHub’s Model Context Protocol (MCP) Registry, a standardized system enabling large language models to interact with external tools and perform actions like coding and testing within VS Code. He highlights the creation of an open-source canonical MCP registry to streamline discovery and management of MCP servers, and envisions future advancements in enterprise governance and dynamic tool composition to enhance security and developer productivity.
In this episode of the VS Code Insiders podcast, host James Montto Magno interviews Toby Padilla, Principal Product Manager at GitHub, about the origins and evolution of the GitHub Model Context Protocol (MCP) Registry. Toby shares his background, including his experience as a founder and engineer, and explains his role in leading GitHub’s MCP initiatives. He highlights the rapid growth and innovation in the MCP ecosystem over the past year, including the launch of the GitHub MCP server and the integration of agentic workflows like the Copilot coding agent.
Toby explains that the Model Context Protocol (MCP) is a standardized way for large language models (LLMs) to interact with external systems by fetching up-to-date or private information and performing actions such as creating repositories or pull requests. MCP servers expose these capabilities as functions that LLMs can call. He describes how GitHub’s MCP server is one of the most popular reference implementations and how it integrates with tools like the Copilot coding agent and Microsoft’s Playwright MCP server to enable advanced workflows such as automated coding and end-to-end testing directly from within VS Code.
The conversation then shifts to the challenge of discovering and managing the growing number of MCP servers and registries. Toby outlines how the MCP ecosystem initially saw a proliferation of independent registries, which created difficulties for publishers and users. To address this, the MCP steering committee developed an open-source canonical MCP registry that acts as a single source of truth for MCP server metadata. This registry allows publishers to submit their MCP servers once, and multiple downstream registries—including GitHub’s MCP registry—can sync from it, enhancing the data with platform-specific information like star counts and installation options.
James and Toby discuss the practical aspects of publishing MCP servers, clarifying that the registry does not host binaries but rather metadata and pointers to where MCP servers are hosted, such as Docker Hub or npm. Toby explains that publishers use a command-line tool to create a manifest file describing their MCP server, including how to run it locally or access it remotely, and then publish this manifest to the open-source registry. This approach supports a federated ecosystem where different platforms can tailor the registry data to their users’ needs, including enterprise scenarios where governance and security are critical.
Looking ahead, Toby envisions two major areas of growth for MCP registries: enterprise governance and dynamic tool discovery. Enterprises will benefit from allow-listing and governance features that ensure only trusted MCP servers are used internally, addressing security and compliance concerns. On the innovation front, registries could enable agents to dynamically compose virtual MCP servers by combining tools from multiple sources to fulfill complex tasks. This vision points to a future where MCP registries not only catalog tools but actively empower intelligent agents to discover, install, and use the best tools on demand, streamlining developer workflows and enhancing productivity.
