How to build an agent that does what you want without writing any code. • This is open-source, so you can host it anywhere • Vendor-neutral, so you can use any model you want • You can set the entire agent loop without writing any code See the GitHub... Repository below.show more

Check out their GitHub repository and give it a star if you find it useful:

@svpino sounds interesting! curious how it handles complex tasks without custom coding. gonna check it out.

Boom, let’s freaking go!!

Such a clear walkthrough- excited to build and use my own agents!

the compaction story makes sense for cost. whats the failure mode when it trims something the agent needed 20 turns later? curious if solve rate holds on tasks longer than the benchmark ones

no code agents shipping faster than anyone can debug them

might wanna mark it a paid post, it is a paid post

Yes. Open source and vendor-neutral remove the lock-in. I see teams here in the States run these loops on any model they choose. More on it:

Or try and build your agent super simple.

Very cool my only initial concern about lock-in is Daytona being the only supported sandbox provider currently. No built-in support yet for a pure local/self-hosted sandbox.

2026 and now we got “agent harnesses” running “agents” that execute “agent loops”. at some point were just gonna sell abstractions wrapped in abstractions with nothing inside

无代码搭代理是省事，难的是把任务边界说清楚。