Ask HN: How are teams sharing AI/agent setups internally?
TL;DR
A Hacker News thread asks how teams manage and distribute shared AI and agent setups internally. In the case described, context, skills, workflows, MCPs and connections to internal tools sit in a local setup, with GitHub serving as the single source of truth. Two pain points keep coming up: onboarding less technical users through Git, cloning, auth and local config, and keeping everyone in sync whenever one person updates the shared context.
Nauti's Take
The encouraging part is that running shared AI setups is finally treated as its own problem: a central configuration source with versioning and automatic sync saves the whole team a daily round of Git rituals. The limit shows up at onboarding, since permissions, secrets and rollbacks get riskier with every shared setup.
Small teams should start with one shared source before building an internal platform.
Sources