Anthropic’s Katelyn Lesse and Angela Jiang lead the team building the company’s developer platform, which serves as both the external API layer for applications accessing Claude’s intelligence and the internal infrastructure for Anthropic’s own apps. Angela frames the platform as a three-layer stack: knowledge, execution, and coordination. She says the real leverage is at the top, in “strategies,” or meta-harnesses that assign different jobs to tokens, such as advising, executing, reflecting, or memory. On the open ecosystem versus walled garden question, they say they are not precious about owning the stack, instead betting on standards by handing skills and MCP to the whole industry, building connectors on the MCP spec, and helping agents—both Claude and non-Claude—work together. The one area they keep closed is model routing, arguing harnesses should be tuned to a model family rather than routing across models. Their frame for the ecosystem is electricity: transformative because everyone could plug in, and no company wired it alone. Angela notes that the coordination layer involves composing orchestrated strategies that sit on top of execution, and that their roadmap will move from the knowledge layer to the execution layer and then to the coordination layer in terms of abstractions they release.