← back
MCP: Origins and Requests For Startups — Theodora Chu, Model Context Protocol PM, Anthropic
Takeaway
MCP went from internal Anthropic hack to multi-lab standard by optimizing for model agency, server simplicity, and bidirectional agent communication.
Summary
- Theo Chu (Anthropic PM) traces MCP origins to co-creators David and Justin in Nov 2024 hack week, motivated by constant copy-paste of Slack/Sentry context and the need for model 'agency'.
- Adoption inflections: Cursor's adoption sparked coding-IDE wave, followed by VS Code, Sourcegraph, and recently Google, Microsoft, OpenAI also signing on.
- Design choices include streamable HTTP transport (over SSE) for bidirectionality, optimizing server simplicity at the cost of client complexity, fixing OAuth in draft spec via community PRs.
- Roadmap and RFS: elicitation (servers asking users for clarification), registry, agent-graph patterns, with startup opportunities around server tooling, observability, and identity.
mcpanthropicprotocols
Original description
Learn more about the latest updates on MCP and get ideas for what startups to build. About Theodora Chu Theo is a product manager at Anthropic, focused on bringing knowledge to models. She works on the Anthropic API as well as MCP. Prior to Anthropic, she spent much of her career building zero-to-one products at her own startup as well as at Stripe. She's come full circle since dropping out of her master's in NLP at Stanford. Recorded at the AI Engineer World's Fair in San Francisco. Stay up to date on our upcoming events and content by joining our newsletter here: https://www.ai.engineer/newsletter