One Year of MCP — with David Soria Parria and AAIF leads from OpenAI, Goose, Linux Foundation | Latent Space

The Latent Space podcast features a discussion on the first anniversary of the Model Context Protocol (MCP) and the launch of the Agentic AI Foundation (AAIF). David Soria Parra from Anthropic reviews MCP's adoption by major tech companies like Microsoft, Google, and OpenAI, as well as protocol improvements including remote access and authentication. The panel explores the rationale behind housing MCP within the AAIF, emphasizing its neutrality and the desire to foster open standards in the AI industry. Jim Zemlin, Nick Cooper, and Brad discuss the foundation's goals, including IP management, community building, and event hosting, while promoting composability and demonstrated utility in potential projects. They envision a future where agents operate asynchronously, driving innovation and impactful applications across various sectors.

Outlines

Part 1: Introduction, MCP Origins

Part 2: Technical Architecture, Transport

Part 3: Connectivity, Skills, Internal Use

Part 4: Community, Use Cases, Tasks

Part 5: Developer Experience, Security

Part 6: Foundation Governance, Partners

Part 7: Principles, Roadmap, Incentives

Part 8: Future Outlook, Community Impact

Sign in to continue reading, translating and more.

Continue

One Year of MCP — with David Soria Parria and AAIF leads from OpenAI, Goose, Linux Foundation

Latent Space

Part 1: Introduction, MCP Origins

Introduction to MCP, the Agentic AI Foundation, and a One-Year Recap

MCP's Evolution: From Local Servers to Remote Access and Authentication Improvements

David's Role in MCP and the Foundation: Neutrality, Governance, and Authentication

Part 2: Technical Architecture, Transport

Transport Challenges: Streamable HTTP, Bidirectional Streams, and Scalability

Standardization Process: Balancing Consensus, Speed, and Open Source Principles

Model Improvements and Protocol Design: Progressive Discovery and Code Mode

Part 3: Connectivity, Skills, Internal Use

MCP vs. Skills: Connectivity Layer vs. Domain Knowledge

Anthropic's Internal MCP Usage: Dogfooding, Gateways, and Internal Apps

MCP Registries: Discovery, Trust, and Standardization

Part 4: Community, Use Cases, Tasks

Community Insights: Financial Regulations and Creative Use Cases

Sampling Use Cases and MCP's Name

Long-Running Tasks: Enabling Agents and Deep Research

Task Design: Asynchronous Operations and Context Compression

Part 5: Developer Experience, Security

MCP for AI App Developers vs. AI Consumers

MCP Apps: UI, Security, and Financial Services Extensions

Call to Action and Addressing Concerns About the Foundation

Part 6: Foundation Governance, Partners

Introduction to AAIF Core Team Members and the Foundation's Origins

Linux Foundation's Perspective: MCP's Early Stage and Industry Impact

Structuring the Foundation: Governance, Technical Contributions, and Support

Block's Involvement: Goose, Concrete Implementations, and Ecosystem Value

Part 7: Principles, Roadmap, Incentives

Openness, Collaboration, and Roadmap Principles

Composability, Traction, and Established Niches

Incentives for Donating Projects to the Foundation

Technical Contributions, Communication, and Client Value

Part 8: Future Outlook, Community Impact

Why a New Foundation? Focus and Community

Actionable Outcomes: IP Management, Leadership, and Community Engagement

Open Source and Community Influence

Future Expectations: Success Stories and Impactful Technology

One Year of MCP — with David Soria Parria and AAIF leads from OpenAI, Goose, Linux Foundation

Latent Space

Part 1: Introduction, MCP Origins

00:06Introduction to MCP, the Agentic AI Foundation, and a One-Year Recap

Introduction to MCP, the Agentic AI Foundation, and a One-Year Recap

01:16MCP's Evolution: From Local Servers to Remote Access and Authentication Improvements

MCP's Evolution: From Local Servers to Remote Access and Authentication Improvements

05:21David's Role in MCP and the Foundation: Neutrality, Governance, and Authentication

David's Role in MCP and the Foundation: Neutrality, Governance, and Authentication

Part 2: Technical Architecture, Transport

11:41Transport Challenges: Streamable HTTP, Bidirectional Streams, and Scalability

Transport Challenges: Streamable HTTP, Bidirectional Streams, and Scalability

15:36Standardization Process: Balancing Consensus, Speed, and Open Source Principles

Standardization Process: Balancing Consensus, Speed, and Open Source Principles

18:31Model Improvements and Protocol Design: Progressive Discovery and Code Mode

Model Improvements and Protocol Design: Progressive Discovery and Code Mode

Part 3: Connectivity, Skills, Internal Use

23:36MCP vs. Skills: Connectivity Layer vs. Domain Knowledge

MCP vs. Skills: Connectivity Layer vs. Domain Knowledge

26:11Anthropic's Internal MCP Usage: Dogfooding, Gateways, and Internal Apps

Anthropic's Internal MCP Usage: Dogfooding, Gateways, and Internal Apps

30:41MCP Registries: Discovery, Trust, and Standardization

MCP Registries: Discovery, Trust, and Standardization

Part 4: Community, Use Cases, Tasks

36:17Community Insights: Financial Regulations and Creative Use Cases

Community Insights: Financial Regulations and Creative Use Cases

39:01Sampling Use Cases and MCP's Name

Sampling Use Cases and MCP's Name

41:11Long-Running Tasks: Enabling Agents and Deep Research

Long-Running Tasks: Enabling Agents and Deep Research

45:28Task Design: Asynchronous Operations and Context Compression

Task Design: Asynchronous Operations and Context Compression

Part 5: Developer Experience, Security

50:11MCP for AI App Developers vs. AI Consumers

MCP for AI App Developers vs. AI Consumers

55:01MCP Apps: UI, Security, and Financial Services Extensions

MCP Apps: UI, Security, and Financial Services Extensions

59:28Call to Action and Addressing Concerns About the Foundation

Call to Action and Addressing Concerns About the Foundation

Part 6: Foundation Governance, Partners

1:03:30Introduction to AAIF Core Team Members and the Foundation's Origins

Introduction to AAIF Core Team Members and the Foundation's Origins

1:06:19Linux Foundation's Perspective: MCP's Early Stage and Industry Impact

Linux Foundation's Perspective: MCP's Early Stage and Industry Impact

1:09:21Structuring the Foundation: Governance, Technical Contributions, and Support

Structuring the Foundation: Governance, Technical Contributions, and Support

1:11:12Block's Involvement: Goose, Concrete Implementations, and Ecosystem Value

Block's Involvement: Goose, Concrete Implementations, and Ecosystem Value

Part 7: Principles, Roadmap, Incentives

1:14:27Openness, Collaboration, and Roadmap Principles

Openness, Collaboration, and Roadmap Principles

1:17:42Composability, Traction, and Established Niches

Composability, Traction, and Established Niches

1:20:58Incentives for Donating Projects to the Foundation

Incentives for Donating Projects to the Foundation

1:22:32Technical Contributions, Communication, and Client Value

Technical Contributions, Communication, and Client Value

Part 8: Future Outlook, Community Impact

1:23:58Why a New Foundation? Focus and Community

Why a New Foundation? Focus and Community

1:27:35Actionable Outcomes: IP Management, Leadership, and Community Engagement

Actionable Outcomes: IP Management, Leadership, and Community Engagement

1:31:44Open Source and Community Influence

Open Source and Community Influence

1:34:52Future Expectations: Success Stories and Impactful Technology

Future Expectations: Success Stories and Impactful Technology