
Rhys
@RhysSullivan • 55,915 subscribers
building https://t.co/gIFOwRUrLU to connect any agent to every tool, API, and service your team uses
Shorts
Videos

I'm now building Executor full time as a startup! The state of tool calling is a mess: - Everyone is using different agents - You have no idea what actions are going to run - Destructive actions are treated the same as normal Executor lets you set up everything once & use it with any agent. You're able to control what tools can be called, require approval on destructive actions, and invite your team to all use them. Under the hood, it's built on codemode to let you add thousands of tools with 0 context bloat. I've been using it for a month so far and it has completely changed how I interact with agents. There's so much to be built here, immediately next up on the roadmap is workspaces, skills, workflows, and generative ui. Everything is being built open source over at there's a lot more coming on an SDK to embed it into your own agents, a self hostable cloud app, and plugins There's an early beta version over at It's very early on - if you hit any bugs let me know and I'll get them fixed up, excited to share more as I build it!
Rhys200,267 次观看 • 26 天前

I'm concerned we're entering a local maxima with CLIs, they're the wrong interface for agents The right interface is regular REST APIs with CIMD (same spec that MCP uses to allow for dynamic client registration) Your agent then writes code to interact with the API (like Cloudflare codemode) Think about how this works for humans today for the following action: "I want to set a DNS record on my domain" -> You Google "Vercel set DNS records" -> Docs page tells you what buttons to press -> You press buttons on a website, those call an API Now for agents, they search Google: -> "Set DNS record on Vercel Domain" -> they land on the same docs page, except it outlines what api endpoints are used as well -> the agent then calls those api endpoints for you, credentials are dynamically inserted where the agent can run them -> (optional) set auto approval policies / require approval of all non GET options by default You don't need separate interfaces for agents, nor do you really need separate skills for them CLIs have terrible discoverability, no input / output typing, they're harder to make profiles for for allowed / disallowed tools They work as a stop gap solution, but companies should be focusing on making good docs and APIs, not CLIs I've been prototyping this over at (open source if you want to play with what this world would look like - still early on it so appreciate feedback - open source and can run completely on your machine
Rhys220,825 次观看 • 3 个月前