GXV - AI Orchestrator
The Problem
Teams running multiple AI coding agents on the same codebase had no way to prevent them from colliding. Two agents would edit the same files, produce conflicting changes, and silently break each other's work. Without visibility into what each agent was doing, developers lost hours to merge conflicts, duplicated effort, and cascading failures — turning a productivity tool into a source of chaos.
Our Approach
We built GolemXV as a coordination layer — not an orchestration framework. Agents check in via REST API, declare their work scope, and instantly learn who else is active in the same area. The system detects overlapping scopes and issues advisory warnings before conflicts happen. Real-time broadcasting through Centrifugo keeps all agents and human operators aware of changes the moment they occur. A live dashboard gives teams full visibility: active agents, conflict history, session timelines, and GitHub issue status — all updating in real time. Integration with GitHub, Mattermost/Slack, and the Claude Agent SDK lets GolemXV fit into existing workflows without replacing them.
The Outcome
GolemXV eliminated blind spots in multi-agent workflows. Teams run 3-5 AI agents concurrently on the same project with zero undetected conflicts. Developers spend their time reviewing agent output instead of resolving merge disasters. The platform coordinates agents across multiple Golem15 projects from a single dashboard, with full audit trails and instant team notifications. What was once chaotic parallel work became structured, observable, and predictable.