This is Hamamoto from TIMEWELL Inc.
In 2026, MCP (Model Context Protocol) has been established as the standard protocol across the AI industry.
In December 2025, the AAIF (Agentic AI Foundation) was established under the Linux Foundation, with Anthropic, OpenAI, Block, Google, Microsoft, and AWS beginning joint operations. MCP has achieved 97 million monthly SDK downloads and over 10,000 active servers, with all major AI platforms including ChatGPT, Claude, Gemini, Microsoft Copilot, and VS Code now supporting it.
This article covers MCP's latest developments, the significance of AAIF's founding, and practical implementation approaches.
MCP 2026 Latest Information
| Item | Details |
|---|---|
| AAIF Founded | December 2025, under Linux Foundation |
| Monthly SDK Downloads | Over 97 million |
| Active Servers | Over 10,000 |
| Supported Platforms | ChatGPT, Claude, Gemini, Copilot, VS Code |
| Platinum Members | AWS, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, OpenAI |
| AGENTS.md Adoption | Over 60,000 OSS projects |
| Gartner Forecast | 40% of enterprise apps with AI agents by end of 2026 |
| MCP Dev Summit | April 2–3, 2026, NYC |
What Is MCP?
Core Concept
MCP (Model Context Protocol) is a standardized protocol for large language models (LLMs) to communicate with external tools and data sources.
MCP's Three Primary Functions:
- Tools: AI executes external actions (file creation, API calls, database operations, etc.)
- Resources: External data the AI can access (files, documents, databases, etc.)
- Prompts: Reusable instruction templates optimized for specific tasks
Why MCP Matters
Traditional AI assistants only referenced conversation history (internal context). MCP standardizes access to "external context," enabling AI to:
- Search for the latest web information
- Reference internal company documents
- Operate task management tools
- Query databases
- Automate browser operations
...and execute these advanced tasks seamlessly.
AAIF's Founding — A Turning Point for Industry Standardization
Agentic AI Foundation
On December 9, 2025, the Linux Foundation announced the establishment of the AAIF (Agentic AI Foundation).
Founding Members:
- Anthropic: Contributed MCP (Model Context Protocol)
- OpenAI: Contributed AGENTS.md
- Block: Contributed goose
Platinum Members
AAIF Platinum members include the major players in the AI industry:
- Amazon Web Services
- Anthropic
- Block
- Bloomberg
- Cloudflare
- Microsoft
- OpenAI
Gold Members: Adyen, Cisco, Datadog, Docker, IBM, JetBrains, Okta, Oracle, Salesforce, SAP, Shopify, Snowflake, and others
Significance of AAIF's Founding
1. Neutral Governance
- Protocol direction is decided by the entire industry, not a single company
- Collaboration from a broad range of participants including competitors
2. Long-term Stability
- Sustainable development as an open standard
- Avoidance of vendor lock-in
3. Cross-Company Interoperability
- A single MCP server built once runs across Claude, ChatGPT, Gemini, and Copilot
- Reduced development costs
Looking for AI training and consulting?
Learn about WARP training programs and consulting services in our materials.
MCP's Rapid Growth
One Year of Results
In just over a year since its release, MCP has become one of the most rapidly growing open-source projects in the AI industry.
Growth Metrics:
- Monthly SDK downloads: Over 97 million
- Active servers: Over 10,000
- Supported platforms: ChatGPT, Claude, Cursor, Gemini, Microsoft Copilot, Visual Studio Code, and many more
OpenAI's Adoption
In March 2025, OpenAI officially adopted MCP, integrating it across their products including the ChatGPT desktop app.
OpenAI's Contribution:
- Early adopter and core contributor to MCP
- Uses MCP as the foundation for ChatGPT's Apps feature
AGENTS.md — Project-Specific Guidance
What Is AGENTS.md?
AGENTS.md is a standard released by OpenAI in August 2025, providing project-specific guidance to AI coding agents.
Features:
- Markdown-based, concise convention
- Standardizes agent behavior across repositories
- Adopted by over 60,000 OSS projects
Adopting Projects and Frameworks: Amp, Codex, Cursor, Devin, Factory, Gemini CLI, GitHub Copilot, Jules, VS Code, and more
MCP Tool Search
Traditional Limitations
In traditional MCP implementations, all tool definitions had to be included in the context window — leading to performance degradation as the number of tools grew.
Tool Search's Innovation
The "MCP Tool Search" feature released in January 2026 enables "lazy loading" of tool definitions.
How It Works:
- AI initially understands only a summary of available tools
- Searches for the tools needed based on the task
- Dynamically loads only the detailed definitions of needed tools
- Releases used tool definitions after task completion
This enables access to thousands of tools.
Use in Claude Code
Integration with MCP
Claude Code is deeply integrated with MCP, enabling the following uses:
1. GitHub Integration
claude mcp add github
> Review this PR and check for any security issues
2. Database Operations
claude mcp add postgres --connection-string "..."
> Write a query to get active users from the users table for the past 30 days
3. Cloud Infrastructure Management
claude mcp add aws
> Check the status of EC2 instances running in this region
Plan Mode
Claude Code's "Plan Mode" achieves its full potential when combined with MCP.
Workflow:
- Claude formulates an execution plan
- Generates an editable
plan.mdfile - Developer reviews and modifies the plan
- Executes using MCP tools after approval
Then vs. Now: MCP's Evolution
| Item | Then (Late 2024, at launch) | Now (January 2026) |
|---|---|---|
| Governance | Anthropic alone | AAIF (under Linux Foundation) |
| Participating Companies | Anthropic-centric | OpenAI, Google, Microsoft, AWS, etc. |
| Monthly Downloads | A few million | Over 97 million |
| Active Servers | A few hundred | Over 10,000 |
| ChatGPT Support | None | Full support (Apps foundation) |
| AGENTS.md | None | 60,000+ project adoption |
| Tool Search | None | Thousands of tools supported |
| Enterprise Adoption | Experimental | Gartner predicts 40% (end of 2026) |
Outlook for 2026 and Beyond
Multimodal Support
In the second half of 2026, MCP multimodal support is planned.
Planned Features:
- Process images, audio, and video via MCP
- Integration with Vision Language Models (VLMs)
- Real-time screen sharing and AI operations
Sampling and Elicitation
Sampling Feature: MCP server side can request inference from LLMs. MCP servers evolve from simple tool providers to bidirectional dialogue partners with AI.
Elicitation Feature: Provides a standardized method for MCP servers to collect additional information from users.
Enterprise-Focused Features
- Standardization of authentication and authorization
- Unified audit log format
- Multi-tenant support
- Rate limiting and quota management
Five Steps to Adopting MCP in Your Organization
Step 1: Identify Use Cases
Identify tasks you want to automate with MCP:
- Internal document search and summarization
- Integration with task management tools
- Report generation from databases
- Customer support automation
Step 2: Leverage Existing MCP Servers
Search the official MCP registry for existing servers. Major services including GitHub, Slack, Notion, and Google Drive already have MCP support.
Step 3: Develop Custom MCP Servers
For use cases that existing servers can't cover, develop custom MCP servers. SDKs are available in multiple languages including TypeScript/Node.js, Python, and Go.
Step 4: Security and Governance
- Access control for MCP servers
- Policies for handling sensitive data
- Audit log configuration
Step 5: Continuous Improvement
Monitor MCP usage and continuously add and improve tools.
Summary
MCP has been established as the industry standard for AI agents through the founding of AAIF.
Key points from this article:
- AAIF (Agentic AI Foundation) was established in December 2025 under the Linux Foundation
- Major companies including Anthropic, OpenAI, Google, Microsoft, and AWS jointly operate it
- Over 97 million monthly SDK downloads, more than 10,000 active servers
- All major platforms including ChatGPT, Claude, Gemini, Copilot, and VS Code support it
- AGENTS.md adopted by over 60,000 OSS projects
- Tool Search enables access to thousands of tools
- Gartner predicts 40% of enterprise apps will have AI agents by end of 2026
- MCP Dev Summit 2026 on April 2–3 in NYC
About one year after its late 2024 launch, MCP has evolved from "Anthropic's proprietary protocol" to "the common language of the AI industry." With AAIF's founding, a structure has been established where even competing companies collaborate to advance the protocol — and developers can now build an MCP server once and use it across any AI platform.
For companies considering AI adoption, MCP support is now essential. We recommend starting small with existing MCP servers, then progressively developing custom servers and enterprise-grade implementations.