Overview
In March 2025, OpenAI officially adopted the Model Context Protocol (MCP), originally introduced by Anthropic, marking a significant industry milestone. CEO Sam Altman announced that OpenAI would add MCP support across its products, including the ChatGPT desktop app.
Announcement and Impact
- March 2025: OpenAI announces MCP adoption
- Industry Validation: Major competitor adopting Anthropic's standard
- Cross-Product Integration: MCP support across OpenAI product line
- Developer Ecosystem: Python SDK and documentation released
OpenAI MCP Features
- ChatGPT Desktop Integration: MCP servers accessible from ChatGPT
- Python SDK: Official OpenAI Agents Python SDK with MCP support
- Tool Discovery: Automatic discovery of MCP tools
- Unified Interface: Consistent integration across OpenAI products
- Agent Framework: MCP as tools for OpenAI agents
Significance
- Validates MCP as industry standard
- Enables cross-platform AI tool compatibility
- Unifies ecosystem previously fragmented by proprietary standards
- Demonstrates collaboration between AI leaders
- Accelerates MCP adoption across industry
Developer Impact
- Single MCP server works with both Claude and ChatGPT
- Reduced integration complexity
- Larger potential user base for MCP tools
- Standardized development patterns
- Cross-platform testing and deployment
Linux Foundation Partnership
In December 2025, Anthropic donated MCP to the Agentic AI Foundation (AAIF), a directed fund under the Linux Foundation, co-founded by Anthropic, Block, and OpenAI, ensuring neutral governance.
Current Scale (2026)
- Over 5,800 available MCP servers
- 97M+ monthly SDK downloads
- Adoption by OpenAI, Google, Microsoft, and others
- De facto industry standard for agentic AI integration
Use Cases
- Universal AI tool development
- Cross-platform AI applications
- Enterprise AI integration
- Developer tool ecosystems
Documentation
Official docs at https://openai.github.io/openai-agents-python/mcp/
Pricing
MCP integration included with OpenAI products. Standard OpenAI pricing applies.
