Show HN: OS layer for running multiple Codex agents in parallel (github.com)

emdash is a lightweight Electron/Node UI that turns the OpenAI Codex CLI into a multi‑agent OS layer: you can spin up many Codex agents in parallel, each isolated in its own Git worktree/branch, monitor their status, review diffs, and open PRs from a single dashboard. It requires Node.js 18+, Git, the Codex CLI (npm or brew) and optionally the GitHub CLI for PR and repo features. Workspaces and projects are persisted locally in an SQLite DB (auto-created per platform) with explicit schemas: a projects table tracking remotes/branches/GitHub connection and a workspaces table that records branch, path, agent_id, status and enforces cascade deletion. The app also includes workspace lifecycle hooks, a planning chat with controlled execution, and a pluggable provider system so other CLI agents (Claude Code, Gemini CLI, aider, Warp) can be run alongside Codex. For the AI/ML community this is practical infrastructure for multi‑agent development and experimentation: Git worktrees give strict isolation for fan‑out tasks and safe diffs/PR workflows, SQLite session persistence enables reproducible local state, and hooks/automation let you run setup/cleanup per workspace. Because data is stored locally, it supports privacy and offline use, but using Codex/GitHub CLIs still transmits data to those providers under their policies. Overall, emdash lowers the friction to orchestrate, compare, and manage multiple coding agents without central servers, making multi‑agent engineering workflows more accessible.