Your teammate’s agent just renamed the auth module. Yours has no idea. Colaberi connects every developer’s local Claude Code or Codex session to one shared project brain — same context, same decisions, and overlapping work flagged before it reaches the diff.
$ npm install -g @colaberi/connectorFree for solo developers · Everyone keeps their own agent subscription · Your raw transcript never leaves your machine
Scroll · the film plays as you go
01 · Connect
One npm command in the repo. Claude Code and Codex sessions keep working exactly as before — the connector just watches.
48 frames · generated for this page, scrubbed by your scroll
Parallel agents are the right model — every developer should have one. The problem is that each runs blind.
Your agent doesn't know your teammate's agent just renamed the auth module. Two agents refactor the same file from different assumptions, and nobody finds out until merge.
The team picked React Query on Tuesday. Your agent knows. Nobody else's does. Every decision gets re-explained to every agent, by hand, forever.
Two agents ship contradictory changes. You find out in the pull request, after both developers burned an afternoon on it.
Everyone keeps the agent, the branch and the subscription they already use.
$ npx colaberi connect
$12 per seat, per month. A seat is a developer whose connector publishes into a team workspace. Team plans open shortly — request access and we will email you the moment they do. Everyone keeps their own agent subscription; Colaberi never resells one.
Solo workspaces can't add team members or extra workspaces — that's what Team is for.
No. Colaberi never shares, proxies or resells an agent subscription. Every developer's prompts run on their own account, exactly as they do today. The connector only reads the local session log.
Your raw transcript stays on your machine. The connector redacts secrets locally — AWS keys, GitHub tokens, sk- prefixes, .env values and your own patterns — then publishes a redacted event stream. The cloud never receives the unredacted session.
Yes, and that's the point. Both are normalised into one event shape, so a team running a mix of the two shares a single project brain. You can have Claude Code on one branch and Codex on another in the same repo.
No. You run one command in your repo and then use your agent exactly as before. The connector attaches to the session automatically and re-scans every ten seconds, so start order doesn't matter.
A developer whose connector publishes into a team workspace, at $12 per seat per month. Solo use is free and unlimited, and you are only billed for seats actually publishing.
Free for solo developers. One command to connect, and nothing about how your team already works has to change.