Rhys

building to connect any agent to every tool, API, and service your team uses | yc 26

[Rhys @RhysSullivan]

Executor is joining the YC S26 batch!

We're building an open source MCP gateway to connect any agent to any service

Your team is constantly spinning up new agents, trying out new tools, wrangling multiple accounts. You need one place to configure everything once, and use them anywhere.

Executor gives you the ability to connect your agents to thousands of tools in a way that keeps you in control. Giving agents access to production services and data should be easy and predictable

Since launching the beta we've:
- Shipped a self-hostable Docker version
- Shipped a desktop app
- Shipped setup by chatting with your agent
- Shipped multi-account support
- Hit 2,000 GitHub stars
- Built a product loved by thousands of users

Truly appreciate all of the early support on it, has been incredible seeing all the ways people use it

There's so much to build here. The past month has been laying the foundation. Now that that's in place, next up is custom tools, generative UI, and workflows.

Setting it up is as easy as asking your agent "help use https://t.co/gIFOwRUrLU". Incredibly excited about what's next, let me know all your feedback!

[Rhys @RhysSullivan]

@reed_barnes Whoa these are great shots - what’d you take them on?

[Rhys @RhysSullivan]

@zeeg @ryanvogel What’s something that you’re dynamic loading vs not?

I like having the dynamic loading separate from the harness since you sort of have to implement that anyways once you get into dynamically loading MCPs that people bring

[Rhys @RhysSullivan]

@kamathematic > exposing them all at once and unnecessarily

What do you mean by this?

[Rhys @RhysSullivan]

Oh yeah I like code mode but code mode solves dynamic tool loading for all models without them having to implement it since the model only has search and execute and those never change which is why I was confused on needing sdks to support dynamic loading

Unless you need a dynamic loading sdk for code mode

[Rhys @RhysSullivan]

@zeeg @ryanvogel What do you do for execution environment then so the model can chain / filter calls?

[Rhys @RhysSullivan]

@zeeg @ryanvogel Sort of, but you could represent it as just bash and CLIs instead, then you’re working with the training data

The models have no problem installing a CLI and immediately calling it which is the same as dynamic tool loading

[Rhys @RhysSullivan]

@zeeg @ryanvogel I don’t really get the need for dynamic loading of tools when you can give the model a `search` and `call` tool and get the same benefits

[Rhys @RhysSullivan]

@Mokronoss @zeeg it should do this already, this is for the clients that don't load the `execute` tool description they only list `execute` and that's all the model sees

[Rhys @RhysSullivan]

one thing executor does poorly here is it only adds `execute` to the tool list, so when you add the MCP, since most clients defer tool descriptions they don't know it has things like linear

there's a fairly easy fix here to implement a `search_{linear|notion}` tool that get added top level, i need to ship that fix but have been focused on fixing more of the fundamentals of the product since i finally figured those out

[Rhys @RhysSullivan]

@tomhaerter @atlasflow feels too much like planetscale

[Ryan Morrissey @ryan_morr]

We have spent weeks making the best cloud coding platform out there

Below is a demo of how easy it is to set up

Ship faster and better using Hoplite

Link to sign up in bio

[Rhys @RhysSullivan]

what i've been playing with for the next version of executor is this framework for apps, an app contains:

- services (things to call)
- tools (fromAPI just registers tools, but custom tools exist as an ejection)
- workflows
- webhooks
- skills

then your agent just writes these small apps to extend itself, also since its just JS you can register whatever you want as tools, i.e wrapping the gh cli to only expose a subset

[Rhys @RhysSullivan]

@abeisleem @executorsh 🫡 more config as code coming soon along with a better management view of organizations

[Rhys @RhysSullivan]

@zeke @jamesqquick honestly the models are good enough these days that could just copy 'help me setup https://t.co/YqTmwo2dnx' and it'd work which is interesting, at least if they have web fetch enabled

[Rhys @RhysSullivan]

fable will back as soon as claude completes the prompt "please figure out why fable access ended and restore it"

[Rhys @RhysSullivan]

@michael_chomsky @mynameisyahia Prefetch on hover doesn’t help here since its client side rendered - you’d have to swap to SSR to get 0 loading states but then you have the complexity of use cache and SSR which the gains are somewhat marginal on for an app that already works

[Rhys @RhysSullivan]

@brian_lovin there's no way that 80% of people can't detect sarcasm

[Rhys @RhysSullivan]

@GaddipatiHarsha isn't this product hunt? or is it more for understanding how people launch things

[Rhys @RhysSullivan]

@PranavBobde @theo that's using it in claude code, not using the blender mcp / some other approach