Orca: An Open-Source Agent Development Environment
AICodeKinggo watch the original →
the gist
Orca is an MIT-licensed development environment that orchestrates multiple CLI agents by isolating tasks into Git worktrees, providing a unified interface for terminal management, browser-based UI testing, and cross-agent task delegation.
Agent Orchestration and Workflow
Orca functions as an Agent Development Environment (ADE) designed to replace fragmented terminal workflows. The core architecture utilizes parallel Git worktrees, allowing multiple agents to operate on different branches of the same repository simultaneously without conflict. Users can fan out prompts to multiple agents to compare outputs or delegate tasks between agents using an orchestration skill. This setup supports over 30 CLI agents, including Claude Code, Codeex, and Grok, by autodetecting tools already installed on the host machine.
Integrated Development Features
Each worktree provides a dedicated environment featuring a Monaco-based editor, terminal panes with WebGL rendering, and a built-in browser with a design mode. The design mode includes an element picker that copies HTML and CSS directly to the agent, and a screenshot annotation tool for front-end feedback. The environment also includes an embedded iOS simulator that agents can operate to verify UI interactions. For task management, Orca integrates with GitHub, GitLab, Linear, and Jira, allowing agents to read and update tickets directly.
Monitoring and Automation
Orca provides granular visibility into agent activity and resource usage. A status bar tracks real-time usage meters for various LLM providers, including Claude cache expiry timers and rate-limit countdowns. The platform includes an automation engine that functions as a cron-like scheduler for repository maintenance, such as dependency audits and release readiness checks. A companion mobile app allows users to monitor terminal output, receive notifications, and approve agent steps remotely via local network or Tailscale.