I’ve been playing around with running multiple Codex sessions on the same project. The idea was simple: One agent follows the frontend roadmap. Another follows backend tasks. Another updates docs. Another reviews or prepares deploys. In theory, it’s pretty cool. You can give each one its own role, API key and roadmap, then let them work while you do something else. In practice, I ran into a dumb but annoying problem: They don’t know what the other agents are doing. So one agent might start working on something based on old context while another agent has already changed the same area. Or two agents both think they are safe to touch the same repo/resource. So I built a small tool called Availsync: https://availsync.dev It’s basically a coordination layer for agents. Before an agent starts work, it checks in and asks if it can touch a resource. Availsync then returns things like: proceed skip_run claimed observe_only So instead of agents blindly stepping on each other, they get a simple routing decision first. I’m currently testing it with multiple automated Codex setups, but it should also make sense for MCP-based workflows, Claude Code, Cursor agents, docs agents, deploy jobs, etc. It’s still early, but the flow works.

Availsync

Give your coding agents one place to coordinate

Give your coding agents one place to coordinate