team, what do you think about a proper context window breakdown in the chat UI?

right now the indicator shows how full the window is, not what's filling it. thinking of something like: click the context icon to open a modal showing per-category details (system prompt / tools / messages / tool results / files), expandable to see the biggest token hogs + a rough per-turn cost.

tried to build a quick prototype (we can enhance it as per feedback). let me know what you think

Here need guidance from the team - what's the best solution here, cuz changing the model's characteristic leads to undefined behaviour

Hello, I'm Eloy, and I'd like to contribute in the project 👀

Hello everyone ! I'm Yanis. I've been working as a Software Engineer (AI focused) for nearly a year now, I've been dealing with MCPs for the major part of it. I found this gem of a software and it's clear to me just by reading the docs that you guys understand the problems you're solving very well.

I like what you guys are doing, i'd love to contribute despite my relative lack of experience. This project uses a tech stack I am not familiar with, so I will probably spend more time learning than implementing. If you think I could pick up a non-urgent issue to sink my teeth into, i'd love to contribute.

Thanks.

Just a quick question:

what's the typical ETA on review to understand if any code fixes are required or it's good to go? These 3 are pending. Feel free to close them if you decide that issues are not worth fixing, thx

Congrats on raising with 20VC - I used to listen to his pod casts on BART in the mornings 13 years ago. Wow time flies.

Hey everyone! I'm Junyi, based in London. :gb:

I've spent the past 2.5 years as a founding commercial hire at an AI startup, deploying workflows into heavy industries like building materials, chemicals, and healthcare.

I’ve lived the problem Archestra is solving from the other side: watching enterprises become dependent on external vendors to maintain workflows, while the real context and intelligence sits with their own teams the whole time. The second I saw what you’re building with MCP, it clicked — this is what is going to unlock enterprise AI at scale.

Excited to follow along, share an operator’s perspective, and contribute where I can. Huge congrats on the funding news!

Is there a simple issue to start with?

This is my git https://github.com/EloxZ