simfleet lets Claude Code / Codex agents share one Mac's simulators via session-based device claims and slim ~1.3 GB boots
A developer building a React Native app with parallel Claude Code and Codex agents released simfleet, an open-source macOS tool that manages simulator allocation across agent sessions. It solves device collisions, Metro port conflicts on 8081, and memory exhaustion by attributing devices to session IDs, leasing them via claims, and booting slimmed simulators. The tool targets practitioners running multi-agent coding setups on mobile apps.
- simfleet (open source, MIT, macOS-only) routes each Claude Code or Codex session's device requests by session ID, showing in a dashboard and menu-bar icon which session drives which simulator
- simfleet claim holds devices per session; a bundled skill teaches agents to check simfleet agents and never drive a claimed device
- Lanes (worktree + device + leased Metro port) fail rather than steal another session's device, worktree, or port; agents stop only lanes they started
- Native builds are fingerprinted and single-flight, so multiple agents requesting the same build share one
- Slimmed iOS simulators boot at ~1.3 GB versus ~3.8 GB, making 5+ parallel agents realistic
Related
Concord: MCP/CLI for coding agents to coordinate and message
Concord is an MCP and CLI tool that lets coding agents like Claude Code, Codex, and Cursor communicate and coordinate. It allows agents to claim work, see what others are doing, and message each other live, preventing duplicated or conflicting changes. It solves the problem of parallel agents lacking task context, similar to giving a team a shared communication channel.
Concord is an MCP and CLI tool that lets coding agents like Claude Code, Codex, and Cursor communicate and coordinate. It allows agents to claim work, see what others are doing, and message each other live, preventing duplicated or conflicting changes. It solves the problem of parallel agents lacking task context, similar to giving a team a shared communication channel.
iOS Simulator Lease Manager for Coding Agents
A daemon that manages iOS simulators as leases for coding agents, preventing collisions and speeding up agent workflows. It assigns simulators to agents on request, parks idle ones with SIGSTOP to save CPU, and returns warm simulators in ~0.28s instead of a 7s cold boot. Solves the problem of multiple agents fighting over a single simulator, improving reliability and efficiency for developers running parallel coding agents.
A daemon that manages iOS simulators as leases for coding agents, preventing collisions and speeding up agent workflows. It assigns simulators to agents on request, parks idle ones with SIGSTOP to save CPU, and returns warm simulators in ~0.28s instead of a 7s cold boot. Solves the problem of multiple agents fighting over a single simulator, improving reliability and efficiency for developers running parallel coding agents.
Crewmates: Claude Code sessions messaging via hooks
Crewmates is a tool that enables multiple Claude Code sessions working in the same repository to communicate with each other using Claude Code's native hooks feature. It solves the problem of parallel agents conflicting over files and lacking awareness of each other's changes, making multi-agent development more coordinated.
Crewmates is a tool that enables multiple Claude Code sessions working in the same repository to communicate with each other using Claude Code's native hooks feature. It solves the problem of parallel agents conflicting over files and lacking awareness of each other's changes, making multi-agent development more coordinated.
Shikigami: run AI coding agents in parallel, each in a Git worktree
Shikigami is a tool that lets developers run multiple AI coding agents (e.g., Claude, Codex) in parallel, each isolated in its own Git worktree. It solves the problems of not knowing which agent is stuck and agents stepping on each other's changes when sharing a single branch.
Shikigami is a tool that lets developers run multiple AI coding agents (e.g., Claude, Codex) in parallel, each isolated in its own Git worktree. It solves the problems of not knowing which agent is stuck and agents stepping on each other's changes when sharing a single branch.
Show HN: Agentbox – MIT-licensed CLI teleports repos into local or cloud sandboxes for parallel Claude/Codex agents
A developer known as Marco launched Agentbox, an MIT-licensed CLI that copies a repo into Docker or cloud sandboxes to run coding agents like Claude, Codex, and PI in parallel without worktree conflicts. It keeps git credentials and gh tokens on the host, relaying changes through approval-gated shims. The tool targets developers juggling multiple subagents across cloud sandboxes while keeping their existing IDEs and workflows.
A developer known as Marco launched Agentbox, an MIT-licensed CLI that copies a repo into Docker or cloud sandboxes to run coding agents like Claude, Codex, and PI in parallel without worktree conflicts. It keeps git credentials and gh tokens on the host, relaying changes through approval-gated shims. The tool targets developers juggling multiple subagents across cloud sandboxes while keeping their existing IDEs and workflows.