This is Hamamoto from TIMEWELL.
You've probably been hearing the term "MCP" a lot lately.
MCP (Model Context Protocol) is an open standard for connecting AI agents to external tools and data sources. Announced by Anthropic in November 2024, OpenAI, Google, and Microsoft all adopted it in 2025. It now records over 97 million monthly SDK downloads and is being called the "USB standard of the AI agent era."
This article explains how MCP works, where the major players stand on adoption, and how enterprises should approach implementing it.
What Is MCP: The AI Agent Standard
Core Concept
MCP (Model Context Protocol) is a standard protocol that allows AI (LLMs) to interface with external tools, data sources, and systems.
MCP's Role:
- Standardizes the connection between AI agents and external tools
- Like "USB" — any AI can connect to any tool using a common method
- Avoids vendor lock-in
- Secure data access
Why MCP Is Necessary
Previously, integrating AI with external tools required a separate custom integration for each tool. With MCP, a tool needs to be made MCP-compatible just once, and it becomes usable from any AI platform.
Before MCP:
- Develop separate plugins for ChatGPT, Claude, and Gemini
- Implement against each platform's API specification
- High maintenance cost
After MCP:
- Create one MCP server
- Common use across ChatGPT, Claude, Gemini, and others
- Standardized interface
The Evolution of MCP: 2024 to 2026
Key Milestones
| Date | Event |
|---|---|
| November 2024 | Anthropic announces MCP |
| March 2025 | OpenAI adopts MCP (ChatGPT Desktop) |
| April 2025 | Google DeepMind announces MCP support in Gemini |
| May 2025 | Microsoft announces MCP support in Windows 11 |
| November 2025 | MCP Apps Extension (SEP-1865) released |
| December 2025 | Transferred to Agentic AI Foundation under Linux Foundation |
Breadth of Adoption
Status as of January 2026:
- Monthly SDK downloads: 97 million+
- MCP servers: 5,800+
- MCP clients: 300+
- Platinum members: AWS, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, OpenAI
Looking for AI training and consulting?
Learn about WARP training programs and consulting services in our materials.
Major Enterprise MCP Adoption
OpenAI
On March 26, 2025, CEO Sam Altman tweeted: "People love MCP and we are excited to add support across our products" — announcing OpenAI's full adoption of MCP.
OpenAI's MCP Implementation:
- ChatGPT Desktop App: MCP-enabled
- Agents SDK: MCP integration
- Responses API: MCP support
At OpenAI DevDay 2025, "Apps in ChatGPT" was demonstrated — integrations with Figma, Spotify, Booking.com, and other external apps via MCP, showing the ecosystem's expansion.
Google DeepMind's Demis Hassabis has announced MCP support in Gemini models.
Google's MCP-Related Features:
- Gemini Enterprise: Agent integration foundation
- Agentspace: Workflow design environment
- NotebookLM: Enterprise data utilization
Microsoft
At Microsoft Build 2025, GitHub and Microsoft announced participation in the MCP steering committee.
Microsoft's MCP Implementation:
- Windows 11: Native MCP support (preview)
- GitHub: MCP steering committee membership
- Copilot: MCP integration direction
Anthropic
Anthropic, MCP's creator, continues to lead its development.
Anthropic's MCP Use:
- Claude Desktop: MCP standard built-in
- Claude Code: Tool integration via MCP
- Claude Cowork: MCP Skills integration
Agentic AI Foundation: MCP's New Home
Transfer to Linux Foundation
In December 2025, Anthropic donated MCP to the "Agentic AI Foundation (AAIF)" under the Linux Foundation.
AAIF Overview:
- Co-founders: Anthropic, Block, OpenAI
- Platinum members: AWS, Bloomberg, Cloudflare, Google, Microsoft
- Purpose: Promoting open standards for AI agent technology
Projects Donated to AAIF:
- Model Context Protocol (MCP)
- goose (AI development assistant)
- AGENTS.md (agent specification definition)
What It Means for the Industry
MCP's transfer to the Linux Foundation has removed it from any single company's control, establishing it as a genuine industry standard. The fact that OpenAI, Google, and Microsoft — all competitors — are participating in the same foundation is a rare event in tech history.
Technical Features of MCP
H2 2025 Updates
MCP has continued to evolve, with the following features added in the second half of 2025:
Async Operations:
- Previous sequential processing caused wait times
- Parallel processing enables significant speed improvements
- Multiple tasks can execute simultaneously
Server Identity:
- Each MCP server explicitly declares its capabilities
- Understand capabilities before connecting
- Supports appropriate agent selection
MCP Apps Extension (SEP-1865):
- Released jointly by Anthropic and OpenAI in November 2025
- Provides standardized UI capabilities
- Standard interface for app integration
Security Considerations
A security analysis in April 2025 identified several areas to watch.
Key Considerations:
- Prompt injection
- Tool permission management
- External content validation
Security considerations are essential when enterprises deploy MCP.
How Enterprises Can Leverage MCP
MCP Adoption Patterns
Pattern 1: Using Existing MCP Servers
- Leverage 5,800+ existing MCP servers
- GitHub, Slack, Notion integrations available immediately
- No custom development needed
Pattern 2: Building Custom MCP Servers
- Connect internal systems to AI
- Access to proprietary data sources
- Domain-specific tool provision
Pattern 3: Using MCP Clients
- Use Claude Desktop, ChatGPT Desktop, etc.
- Unified interface for tool usage
- Improved end-user productivity
Implementation Steps
- Current State Analysis: Identify the tools and data sources you want to connect to AI
- Survey Existing Servers: Check for publicly available MCP servers
- Custom Development: Build MCP servers as needed
- Security Design: Design access controls and audit logs
- User Rollout: Training and deployment to end users
Enterprise Considerations
Enterprises deploying MCP at scale should consider:
- Governance: Which MCP servers to permit
- Security: Limiting data access scope
- Auditing: Managing logs of AI-tool interactions
- Training: Improving employee MCP literacy
At TIMEWELL, we offer enterprise-grade AI adoption infrastructure through ZEROCK. While leveraging open standards like MCP, ZEROCK provides high-accuracy internal information search via GraphRAG, enterprise-grade security, and knowledge control features.
Our WARP consulting service also supports effective deployment and utilization of AI agent technologies including MCP, with monthly updates to keep you current.
Then vs. Now: MCP's Evolution
Compared to around October 2025 when the original discussions underlying this article took place, MCP has advanced significantly.
Then (around October 2025):
- Anthropic-led
- OpenAI adoption had just been announced
- Practical deployment was limited
Now (January 2026):
- Transferred to AAIF under Linux Foundation
- Major players including OpenAI, Google, Microsoft, and AWS participating
- 97 million+ monthly SDK downloads
- 5,800+ MCP servers
- Established as the de facto industry standard
MCP's Future: 2026 Outlook
Expected Developments
- Accelerated Enterprise Adoption: Full-scale deployment at large enterprises
- Security Enhancements: Enterprise-grade authentication and authorization
- Expanded Use Cases: Healthcare, finance, education applications
- Tool Ecosystem Maturity: More MCP servers and clients
"2025 Was Adoption; 2026 Is Scale"
The industry saying is: "If 2025 was the year of MCP adoption, 2026 is the year of scale." MCP will continue to evolve as the standard infrastructure for AI agents.
Summary
MCP is advancing as the standard protocol for connecting AI agents to external tools, with adoption accelerating across the entire industry.
Key Points:
- MCP is a "USB-like" standard for connecting AI and tools
- Adopted by OpenAI, Google, Microsoft, AWS, and other major players
- Transferred to AAIF under Linux Foundation, establishing it as a neutral standard
- 97 million+ monthly SDK downloads, 5,800+ MCP servers
- Enterprises can leverage existing MCP servers or build custom ones
In the age of AI agents, MCP is becoming a standard no organization can afford to ignore. We recommend building a thorough understanding of MCP early and exploring how to integrate it into your AI adoption strategy.
Related Articles
- The Reality of a Part-Time Employee Who Worked Full-Time, Took Two Maternity Leaves, and Changed Her View of Work | TIMEWELL
- Before Paternity Leave — What You Absolutely Must Do to Take Leave Even During a Busy Period
- Pursuing a Hands-On Architecture Firm: Finding My Own Way as the 5th Generation of a Construction Company | Fujita Construction
