Keep your AI work moving when a 5-hour limit hits. AgentDeck puts every Claude and ChatGPT/Codex account in one desktop workspace, shows limits and usage at a glance, and lets you continue sessions across accounts or between Claude and Codex without rebuilding context. Work securely from your phone or tablet, use local or approved models, and optionally mask personal or sensitive data and check untrusted content for prompt injection before sending.
Hey Product Hunt 👋
I built AgentDeck because my AI workflow kept breaking at exactly the wrong time: Claude's 5-hour limit would hit while work was still running, switching subscription accounts meant logging in again, and moving to Codex meant starting over or manually rebuilding context.
AgentDeck is the desktop workspace I wanted:
• Connect multiple Claude and ChatGPT/Codex accounts in one app
• See 5-hour, weekly, API, and token usage in one place
• Continue a Claude session on another account while preserving the original context
• Relay work between Claude and Codex using a recent-context summary and explicit transfer warnings
• Use local models and organization-approved APIs through the Claude Code harness
• Reach attached agents from a phone, tablet, or another computer over a secure device connection
• Optionally mask personal and sensitive context locally and inspect untrusted content for prompt injection before it follows the selected model route
One important boundary: AgentDeck does not change a model provider's storage or training policies. Any provider-side opt-out still needs to be configured with that provider.
AgentDeck is free, and this is its first public release. I would especially value feedback from heavy Claude and Codex users: which part of your multi-account or multi-agent workflow breaks most often today?