Agent Cockpit

Agent Cockpit transforms managing multiple coding sessions into a single, cohesive experience. Key capabilities include: • Parallel orchestration of Claude Code and Codex sessions • Real-time monitoring of all agent tool calls • Centralized approval and denial of agent actions • Unified timeline for replaying session history • Local-first data storage for audit and privacy This open-source dashboard provides a `local-first` control room for developers. It addresses the operational challenges of running multiple code-generating agents by consolidating scatterd approval requests, providing a shared timeline of events, and reducing context-switching. You can intercept agent actions with custom hooks, allowing for fine-grained control and risk management. The platform normalizes events from different providers like Claude Code and Codex into a consistent view. A local Node daemon processes and stores all session history in SQLite, ensuring a complete audit trail without transmitting sensitive project context externally. This robust architecture delivers a secure and transparent development environment. Agent Cockpit is ideal for engineering teams and developers who frequently utilize multiple code-generating assistant sessions and require enhanced visibility, control, and auditing capabilities to maintain productivity and code quality. It ensures that running many coding agents doesn't lead to a loss of control.