If you and a teammate each use a coding agent, CoCodeOp gives both agents a shared place to plan work and review changes. Here s a first run you can try with v1.0.
1. Start the Core. Download the full v1.0.0 bundle and follow the setup guide to start Core and PostgreSQL with Docker Compose on Linux amd64. Give owners an HTTPS address, or have each open a private SSH tunnel.
2. Connect two leads. Create a workspace, invite your teammate, and install the CoCodeOp plugin in each owner s Codex or Claude Code. Ask both leads to call office_bootstrap. They should report the same workspace ID and separate owner and agent IDs.
3. Run a worker on each machine. Each owner starts a Linux runner with rootless Podman and configures an approved Codex or OpenCode worker profile. A Windows owner can keep the desktop client on Windows and run the Linux runner in a local VM. Ask each lead to create a named worker, give it standing instructions and run a small task. Match the run ID from the lead to the run directory on that owner s machine.
Hey Product Hunt, I’m Mikhail. I wanted my Codex and a teammate’s Codex to be able to plan a shared change, hand it to their own local workers and agree on what was ready to ship.
CoCodeOp gives each person a lead that connects over MCP. The leads can create named workers, give them standing instructions, assign tasks and check what’s running. Workers run on their owner’s Linux machine and read and edit the same checkout within granted permissions.
For v1.0 we ran independent leads in two Debian VMs with real Codex and OpenCode workers. The two teams coordinated overlapping edits, ran tests against source snapshots and recovered the accepted checkpoint after a restart.
It’s open source under Apache 2.0. Start with the setup guide and try a small shared task with a teammate.
If your team already uses coding agents, what coordination problem would you want us to solve first? I’m particularly interested in how you handle overlapping edits and API changes across repos.