Last updated: May 25, 2026
CodeGrid helps you run many AI coding agents at once. It is deliberately a thin, transparent layer between you and the agents you choose — never a model, a proxy, or a place your code goes to be stored. Here is how we think about AI.
CodeGrid is an orchestration layer, not an AI provider. It launches the coding agents you already use — Claude Code (Anthropic), Codex (OpenAI), Gemini CLI (Google), and Cursor — each running as its own process. ZipLyne does not train models, fine-tune on your data, or operate any inference of its own.
You install and authenticate each agent with your own subscription or API credentials. CodeGrid stores none of those credentials and adds no model access of its own. You choose which agent and which model runs in each pane.
When an agent works, its requests flow directly from that agent's CLI to its provider under that provider's terms and privacy policy. CodeGrid does not proxy, inspect, log, or store your prompts, responses, or source code. Nothing is sent to ZipLyne.
CodeGrid is designed for human oversight of many agents at once. Its attention detection surfaces panes that are waiting on you — approvals, confirmations, and yes/no prompts — so an agent doesn't act unattended just because you were looking at another pane. Approvals remain yours to give.
Because CodeGrid is open source under MIT, you can audit exactly what it does with agent output and the filesystem. We'd rather you verify than take our word for it.
Your use of each agent is governed by that provider's terms and data policies: Anthropic, OpenAI, Google, and Cursor. Review them to understand how your code is handled by the models you choose to run.