Rhys Sullivan founded Executor, an integration layer that gives AI agents a shared connection to external tools. His projects span agent infrastructure, web applications and earlier gameplay engineering.[2]
Start with Executor →Background
Sullivan’s portfolio records Fortnite gameplay optimization in May through August 2020 and Warzone gameplay systems in May through August 2021. Tasks included slow-function analysis, bug fixes and the red-door hallway system. His recent projects bring that practical engineering focus to web and agent infrastructure. Y Combinator lists Executor in its Summer 2026 batch with Sullivan as founder.[1][2]
Career
Sullivan’s own timeline dates Executor to February 2026 onward and integrations.sh to May 2026 onward. The latter catalogs API, MCP, GraphQL and command-line integration surfaces, including authentication guidance. Earlier entries include a Vercel platform template with sandbox execution and deployment, a browser MCP transport that uses an existing authenticated browser session, and Typelytics, a project for carrying type information from analytics queries to rendered results.[1]
Contributions
Executor puts authentication and tool access in a shared layer between an AI agent and outside services. Instead of connecting each agent separately to Gmail, GitHub or another API, a user connects the account once and makes that connection available to their agents. The product accepts Model Context Protocol servers, OpenAPI descriptions, GraphQL and custom functions. Its current site also describes running scheduled jobs and webhooks, allowing deterministic work to continue without an agent staying active.[3]
Tools, writing & talks
Tools
Start here
Executorexecutor.shShared authentication and integrations for agents, with hosted and local options.
Begin with Executor’s examples to decide whether sharing connections solves your agent setup. The concepts are understandable without code; installation needs an agent client.
Founder.[2][3]Original source- integrations.shintegrations.sh
Registry of integration specs across MCP, OpenAPI, and GraphQL that AI agents use to discover and connect to services.
Catalog workOriginal source
Writing, research & talks
A thread/article by Rhys Sullivan analyzing why MCP (Model Context Protocol) didn't gain traction when it launched alongside early models like Sonnet 3.5 and GPT-4o, and what comes next for agent-to-tool connectivity. Written by a YC-backed founder building infrastructure to connect AI agents to tools, APIs, and services.
Catalog workOriginal source- Rhys Sullivan’s project timelinerhyssul.com
A chronological guide to web, agent and game projects with descriptions and links.
Personal portfolio.[1]