Executor sits in the integration layer between AI agents and the tools they need to call. The official site describes it as an MCP gateway: Claude Code, Cursor, Codex, or any other MCP-speaking client points at one endpoint, while Executor connects the agent to MCP servers, OpenAPI specifications, GraphQL APIs, Google Discovery sources, and custom JavaScript functions. That framing makes it more specific than a general workflow builder and broader than a single MCP server; it is a catalog and routing surface for teams that are accumulating many agent integrations.
The write-time GitHub check for `RhysSullivan/executor` showed 2,269 stars, 140 forks, an MIT license, and a push on 2026-06-24. The README documents `npm install -g executor`, `executor install`, and `executor web`, then explains local background service setup and MCP server usage. The npm package was at version 1.5.19 and exposes an `executor` binary. The product site also describes local desktop and CLI options, Executor Cloud with a free tier, and self-hosting through Docker or a Cloudflare Worker, so the page frames pricing as open-source local software with optional hosted deployment.
Executor's appeal is operational simplicity: instead of configuring the same GitHub, Stripe, Sentry, Cloudflare, or internal API tools separately for every agent, the team can centralize the integration catalog and expose it through MCP. The risk is also obvious: an integration gateway may hold credentials, schemas, and function access that deserve security review. This page should therefore describe Executor as a promising MCP integration layer, not as a proven enterprise security boundary, and teams should validate auth, logging, and deployment mode before placing high-privilege tools behind it.
