Atm I can workaround this by deploying remote MCPs myself. But I think having MCP servers deployed across different clusters and connecting them using something like a tunnel (for example, NetBird) would be really nice
Innokentii Konstantinov (archestra team) —
Hi! Yes, it's something in our plans, not in immediate backlog though. What's your use case? Yes, please, open an issue!
Mergen Orynbayev —
Hmm, the more I think about it the less I'm sure if multi-cluster is the right thing to use. Most of my use cases (cross cluster observability, alerts, metrics and logs) are better solved another way and there are some legal questions about transferring data between AWS EU and AWS US GovCloud. I will start with a single cluster first and see what hurts.
Innokentii Konstantinov (archestra team) —
What's your use case?
Mergen Orynbayev —
Initially I wanted to implement a cross cluster observability. Something like an internal agent/slack bot
Innokentii Konstantinov (archestra team) —
It depends on where your data lives. For example if you have several Grafana instances holding data from various clusters, you can install two mcps pointing on them and prompt agent where to use which one.
Mergen Orynbayev —
Yes, that's what I plan to do. After looking into this a bit more, it turns out it's not that easy to connect to multiple clusters 🙂
Innokentii Konstantinov (archestra team) —
If you need k8s mcp we also have it in our catalog, it should support several kubeconfigs
Mergen Orynbayev —
Yes, I use it already. I will start with one cluster for now and let's see how it goes. Thank you!
Hey team! 👋 I’ve opened a PR to add support for Anthropic Workload Identity Federation (WIF), which enables keyless authentication for Anthropic API access and updates the model fetcher logic.
It includes unit tests, documentation updates, and a quick demo video in the description. Whenever you have a moment, I’d appreciate a review!
Hey everyone! My name is Vadim. Getting back to work after a gap year. I wanna make a few contributions to learn the codebase. Perhaps https://github.com/archestra-ai/archestra/issues/4731 is a good tiny bug to get started with the env (i will do the onboarding and leave a comment there)? p.s. saw your article on fighting ai-prs, good stuff
I'd love to work on #3859 - adding JSON array support to the MCP server arguments textarea.
Already looked at the code (mcp-catalog-form.utils.ts) and have a clear idea of what needs to change. Plan is to auto detect JSON array input and fall back to the existing one per line behavior so zero breaking changes.
Just posted /attempt on the issue. Would appreciate being assigned so I can submit the PR.
Atm I can workaround this by deploying remote MCPs myself. But I think having MCP servers deployed across different clusters and connecting them using something like a tunnel (for example, NetBird) would be really nice
Innokentii Konstantinov (archestra team)9:30 AMOpen in Slack
Hi! Yes, it's something in our plans, not in immediate backlog though. What's your use case? Yes, please, open an issue!
Hmm, the more I think about it the less I'm sure if multi-cluster is the right thing to use. Most of my use cases (cross cluster observability, alerts, metrics and logs) are better solved another way and there are some legal questions about transferring data between AWS EU and AWS US GovCloud. I will start with a single cluster first and see what hurts.
Innokentii Konstantinov (archestra team)3:01 PMOpen in Slack
Initially I wanted to implement a cross cluster observability. Something like an internal agent/slack bot
Innokentii Konstantinov (archestra team)3:06 PMOpen in Slack
It depends on where your data lives. For example if you have several Grafana instances holding data from various clusters, you can install two mcps pointing on them and prompt agent where to use which one.