Show HN: FleetCode – Open-source UI for running multiple coding agents (github.com)

FleetCode is an open-source desktop terminal app that lets you run multiple CLI coding agents (e.g., Anthropic’s Claude, OpenAI Codex) in parallel, with each agent isolated in its own git worktree. Sessions persist across restarts and can be automatically resumed (--session-id <uuid> for new Claude sessions, --resume <uuid> for reopen), and you can rename, close, or delete sessions with automatic worktree cleanup. The UI supports terminal theming, font and cursor settings, pre-run shell setup commands, and integration with Model Context Protocol (MCP) servers (stdio and SSE) to extend agent capabilities. This matters for AI/ML practitioners because it provides safe, reproducible, and concurrent experimentation: each agent operates in an isolated branch-level workspace so file-side effects don’t collide, and sessions are reproducible across restarts. Tech requirements are straightforward (Node.js 16+, git, and the relevant CLI client such as @anthropic-ai/claude-cli or Codex), and FleetCode spawns a new worktree when you choose a project repo and parent branch. Practical notes: macOS users may need to run xattr -cr /path/to/FleetCode.app to clear quarantine, and Claude Code users should disable “Auto connect to IDE” (claude config set autoConnectToIde false) if the agent reads/writes outside the worktree. Overall, FleetCode streamlines multi-agent workflows with isolation, persistence, and flexible integration points.