Agentic AI Foundation: Microsoft, Google, OpenAI and Anthropic join forces for open standards

In an unprecedented move, the biggest competitors in AI have decided to collaborate. Microsoft, Google, OpenAI, Anthropic and other leading companies have formed the Agentic Artificial Intelligence Foundation, managed by the Linux Foundation.

Why this alliance matters

For the first time, companies that fiercely compete in the AI market are working together on open standards. This means:

Before	Now
Each company with proprietary protocols	Shared, open standards
Incompatible AI agents	Cross-platform interoperability
Vendor lock-in	Freedom to switch providers
Ecosystem fragmentation	Unified ecosystem

The 3 key projects

1. MCP (Model Context Protocol) - Anthropic

The Model Context Protocol is an open standard that defines how AI agents connect to external applications.

┌─────────────────────────────────────────────────────────┐
│                    MCP Architecture                      │
├─────────────────────────────────────────────────────────┤
│                                                          │
│   ┌─────────┐      ┌─────────┐      ┌─────────────────┐ │
│   │ Claude  │      │  MCP    │      │ Your Application│ │
│   │ ChatGPT │ ←──→ │ Server  │ ←──→ │ Database        │ │
│   │ Gemini  │      │         │      │ APIs            │ │
│   └─────────┘      └─────────┘      └─────────────────┘ │
│                                                          │
│   Any compatible    Standard        Any external        │
│   AI                protocol        service             │
│                                                          │
└─────────────────────────────────────────────────────────┘

Benefits for developers:

• Single protocol to connect any AI to your services
• Switch AI providers without rewriting integrations
• Open source community developing connectors

2. Agents.md - OpenAI

Agents.md is a standard format for giving instructions to coding agents. It works similar to robots.txt but for AI:

# agents.md example

## Permissions
- Can read files in /src
- Can create files in /src/generated
- Cannot modify /config

## Instructions
- Follow existing code conventions
- Run tests before commits
- Don't expose secrets in logs

## Context
- Framework: Next.js 14
- Testing: Jest + Playwright
- Style: ESLint + Prettier

Benefits:

• Granular control over what an AI agent can do in your repo
• Living documentation of project conventions
• Security by default (principle of least privilege)

3. Goose - Block

Goose is an open source AI agent developed by Block (formerly Square) that will be adopted as the reference implementation:

Feature	Description
Open source	Fully open code
Extensible	Plugin architecture
Multi-model	Works with any LLM
Local-first	Can run completely locally

Implications for businesses

Immediate opportunities

1. Adopt MCP now: If you’re building AI integrations, use MCP from the start
2. Create agents.md: Define rules for AI agents in your repositories
3. Evaluate Goose: As an open source alternative to proprietary agents

Market changes

Before the alliance:
┌─────────┐  ┌─────────┐  ┌─────────┐
│ OpenAI  │  │ Google  │  │Anthropic│
│ Plugins │  │Extensions│ │  MCP    │
└────┬────┘  └────┬────┘  └────┬────┘
     │            │            │
     ▼            ▼            ▼
   Silos       Silos        Silos

After the alliance:
┌─────────────────────────────────────┐
│          Open Standards             │
│     MCP + Agents.md + Goose         │
└──────────────────┬──────────────────┘
                   │
     ┌─────────────┼─────────────┐
     ▼             ▼             ▼
┌─────────┐  ┌─────────┐  ┌─────────┐
│ OpenAI  │  │ Google  │  │Anthropic│
└─────────┘  └─────────┘  └─────────┘

For technical leaders

1. Risk reduction: Open standards = less lock-in
2. Safe investment: Building on MCP is a long-term investment
3. Talent: Look for developers with agentic AI experience

The bigger picture: 2026 is the year of agents

This alliance confirms what analysts predict: 2026 will be the year of agentic AI. AI agents will move from impressive demos to real productivity tools.

“AI agents will proliferate in 2026 and play a bigger role in daily work, acting more like teammates than tools.” — MIT Technology Review

Recommended next steps

Priority	Action	Resource
High	Read MCP documentation	modelcontextprotocol.io
High	Create agents.md in critical repos	Format in development
Medium	Evaluate Goose for internal use cases	Block’s GitHub
Medium	Train team on agentic AI	Specialized courses

---

Want to implement AI agents with open standards in your company? Contact us for an architecture assessment.