Anthropic’s Role in Creating MCP
Overview of MCP (Model Context Protocol)
The Modular Commerce Protocol (MCP) actually refers to the Model Context Protocol – an open standard for connecting AI assistants (LLM-based agents) with external data sources, services, and tools. Think of MCP as a “USB-C for AI” that provides a universal interface for AI models to plug into business systems. By replacing brittle, one-off integrations with a standardized contract, MCP lets AI agents securely access databases, APIs, and e-commerce platforms in real time. This is crucial in e-commerce, where AI-driven agents (for shopping, customer service, etc.) need up-to-date context like product catalogs or order data. The MCP specification and SDKs are openly available (e.g. on GitHub), and it’s being developed as a collaborative open-source project.
Anthropic’s Role in Creating MCP
Anthropic (an AI research company) spearheaded the creation of MCP and remains a core driving force behind the initiative. In November 2024, Anthropic announced it was open-sourcing the Model Context Protocol as a new industry standard to bridge AI assistants with the “systems where data lives”. In other words, Anthropic introduced MCP to tackle the challenge of giving AI models secure, two-way access to enterprise data (content repositories, business apps, etc.) in a consistent way. This has been confirmed by independent sources: IBM, for example, notes that “MCP, introduced by Anthropic in 2024, establishes an open standard for AI–tool interactions”. Anthropic continues to steward the MCP project (which is governed openly – see Governance below) and has built MCP support into its products (e.g. Claude AI’s desktop app can act as an MCP client/server). In summary, yes – Anthropic is not only “part of” the MCP initiative, it originated and actively promotes it.
Major Contributors and Supporters
Beyond Anthropic, a growing ecosystem of companies and organizations is contributing to or adopting MCP, especially in commerce contexts:
Block, Inc. (Square) – An early adopter of MCP. Block’s CTO has praised open protocols like MCP as “bridges that connect AI to real-world applications”, and Block integrated MCP into their systems to build AI “agentic” automation. This indicates fintech and commerce players see value in MCP’s open approach.
Apollo – Listed by Anthropic as an early MCP integrator (likely Apollo GraphQL or a similar platform). By adopting MCP, Apollo enables AI agents to query and interact with data through standard endpoints instead of custom connectors.
Developer Tool Companies – Several AI/DevOps tool providers are actively working with MCP to enhance their platforms. Notable examples include:
Zed (code editor)
Replit (online IDE)
Codeium (AI code assistant)
Sourcegraph (code search engine)
These companies are building MCP compatibility so that AI agents (like coding assistants) can retrieve context (code, documentation, etc.) via a common protocol. Their involvement suggests a broad, cross-industry support for MCP’s standard.
Commercetools – A leading e-commerce platform (and MACH Alliance member) that has embraced MCP to make commerce systems “agent-ready.” In mid-2025 Commercetools introduced “Commerce MCP” as an MCP-based interface to its APIs (carts, products, orders, etc.). Commerce MCP exposes those retail functions in an AI-consumable format, allowing autonomous shopping or analytics agents to safely interact with the platform. According to Commercetools, this provides “standardized contracts for consistent behavior, enabling [AI] agents to understand and interact with commercetools through a unified interface.”. In essence, Commercetools is extending its headless commerce APIs with MCP so AI assistants (like shopping bots or developer copilots) can plug in without custom integration.
Particular Audience – An AI retail technology company that launched Retail-MCP.com in 2025 as a practical demonstration of MCP for online commerce. This initiative provides open-source MCP server connectors for common retail systems and data, allowing AI agents to perform e-commerce tasks (searching products, checking out, managing inventory, etc.) via direct API calls. Particular Audience is “championing open standards like MCP” to help retailers adopt AI agents without sacrificing control. They even offer a pilot program (starting June 2025) to help retailers set up MCP servers and be “agent-ready” for voice commerce, dynamic pricing, automated reordering, and more. This shows that specialized commerce-tech firms are actively driving MCP’s adoption in retail.
IBM – While not an originator, IBM has shown strong support for MCP in enterprise AI contexts. IBM’s AI engineering teams describe MCP as a key standard for agent-tool communication and have built MCP client capabilities into their products (e.g. IBM’s BeeAI agent framework can function as an MCP client). IBM’s developer guides explicitly explain MCP and how it complements their AI orchestration (IBM calls MCP the “HTTP layer for AI” enabling “plug-and-play” tool integrations). This indicates large enterprise vendors view MCP as important for AI integration.
Google and the Linux Foundation – Google has developed a complementary standard called Agent2Agent (A2A) for inter-agent communication, and importantly, Google’s AI teams position A2A as working in tandem with Anthropic’s MCP. In August 2025, the Linux Foundation even welcomed a new project (AGNTCY, backed by Google, Cisco, Oracle, Dell, Red Hat, etc.) aimed at AI agent infrastructure that “interoperates with standards like Agent2Agent (A2A) and Anthropic’s Model Context Protocol (MCP)”. While not directly an e-commerce alliance, this shows a consortium of major tech companies endorsing MCP as part of the emerging open AI stack. It’s likely that as MCP matures, standards bodies or alliances may formally support it; for now, the support is via open-source communities and cross-industry collaborations rather than a single dedicated consortium.
Aside: Notably absent from direct involvement (so far) is OpenAI – instead of MCP, OpenAI has its own plugin system for tools. However, MCP’s open nature means any platform (OpenAI included) could leverage MCP connectors. In fact, MCP servers can even wrap calls to external LLM APIs (OpenAI, etc.) as tools. The key point is MCP is vendor-neutral, so it’s attracting a broad coalition of AI companies, cloud providers, and commerce tech firms even if no single formal “MCP Consortium” exists.
Technical Frameworks and Governance Model
MCP is being built in the open, with a clear technical framework and governance structure:
Open Protocol & SDKs: The MCP specification is openly published (see modelcontextprotocol.io) and reference implementations are on GitHub. Anthropic provided official SDKs in multiple languages to simplify adopting the protocol. There is a growing library of open-source MCP servers (connectors) for popular systems – e.g. Slack, GitHub, Google Drive, databases, etc. – all under permissive licenses. This open-source approach allows any e-commerce platform or tool vendor to create MCP integrations. (For example, Particular Audience’s Retail-MCP connectors are released under open source licenses for retailers to self-host.)
Complementary to Existing Frameworks: MCP is not a standalone “agent platform” but rather a common interface layer. It was designed to work with popular AI frameworks and standards. For instance, MCP can plug into agent orchestration libraries like LangChain, Microsoft’s Copilot stack, or IBM’s orchestration tools – providing the standardized tool/API interface while those frameworks handle agent logic. Similarly, MCP is complementary to Google’s A2A (which standardizes inter-agent messages) – in fact commercetools notes that their Commerce MCP “complements other protocols such as A2A” to cover different layers of the AI stack. In summary, MCP is becoming a unifying layer that many frameworks and cloud providers can support for tool access, rather than competing with them.
Governance and Working Groups: The development of MCP is guided by an open community governance model (inspired by open-source projects like Python or Linux). There is an MCP Steering Group of maintainers and core maintainers drawn from different organizations, with a formal process for proposals (Specification Enhancement Proposals, etc.). Notably, no single company has exclusive control – governance is individual-based, meaning “no seats reserved for specific companies,” to ensure decisions serve the community’s interest. Two lead maintainers (currently Justin Spahr-Summers and David Soria Parra) act as benevolent project leads, but all significant changes go through community discussion and consensus via channels like a public Discord and GitHub issues. This open governance model encourages broad participation from vendors, retailers, and developers alike.
Industry Alliances: While there isn’t a dedicated “MCP Consortium” by name, MCP’s evolution is influenced by industry alliances advocating open, modular commerce and AI standards. For example, the principles of the MACH Alliance (which commercetools co-founded) – API-first, headless, interoperable commerce – align with MCP’s goals of modular integration. We’re also seeing standardization forums (like the Linux Foundation’s AI initiatives) include MCP in their scope. This suggests that MCP is on track to become a widely recognized standard, governed in the open with input from many companies rather than a proprietary spec.
Official Resources and Documentation
For those interested in the MCP initiative, several official resources are available:
MCP Specification and Docs: The Model Context Protocol homepage at modelcontextprotocol.io provides the full specification, developer documentation, FAQs, and community links. This is the best starting point to understand the technical details and how to implement MCP servers/clients.
GitHub Repositories: The core spec and SDKs are on GitHub (linked from the MCP site), along with open-source connectors. Developers can review the code, contribute enhancements, or propose new integrations via this repo.
Anthropic Announcement: Anthropic’s November 2024 announcement post is an official introduction to MCP. It outlines the vision and lists early partner companies, giving insight into the collaborative nature of the project.
Retail-MCP by Particular Audience: The Retail-MCP.com site (announced April 2025) is an official portal for MCP in retail. It offers documentation on deploying retail-focused MCP servers and how to join their pilot program – a useful resource for e-commerce developers.
Commercetools Documentation: Commercetools provides docs for Commerce MCP on their site. This shows a concrete example of MCP applied in a commerce platform, including available APIs and usage guides for commerce scenarios.