Emdash — Parallel Multi-Agent Dev Environment

What it is

Emdash is an open-source agentic development environment (YC W26) that runs multiple AI coding agents in parallel. Each agent operates in an isolated git worktree, preventing conflicts between concurrent code changes. Emdash supports Claude Code, Codex, Gemini CLI, and over 20 other AI coding agents.

Emdash targets engineering teams that want to parallelize AI-assisted development. Instead of running one agent at a time, you can assign different tasks to different agents and merge results.

How it saves time or tokens

Serial AI coding means waiting for one agent to finish before starting the next task. Emdash lets you run five or ten agents simultaneously, each working on a separate feature or bug fix in its own git worktree. The isolation prevents merge conflicts during development.

By distributing tasks across agents, Emdash reduces wall-clock time for large refactoring or feature development projects from hours to minutes.

How to use

1. Install Emdash:

npm install -g emdash

1. Initialize in your project directory:

cd your-project
emdash init

1. Launch parallel agents with task assignments:

emdash run --agent claude-code --task 'Add user authentication'
emdash run --agent claude-code --task 'Write unit tests for the API'
emdash run --agent codex --task 'Refactor the database layer'

1. Each agent works in an isolated git worktree. When done, review the changes and merge them into main.

Example

# Check status of all running agents
emdash status

# Output:
# Agent 1 (claude-code) - auth-feature    [running] worktree: .emdash/wt-1
# Agent 2 (claude-code) - api-tests       [done]    worktree: .emdash/wt-2
# Agent 3 (codex)       - db-refactor     [running] worktree: .emdash/wt-3

# Merge completed work
emdash merge wt-2

Related on TokRepo

• AI Tools for Coding — AI coding assistants and development tools
• Multi-Agent Frameworks — Frameworks for orchestrating multiple AI agents

Common pitfalls

• Running too many agents on a single machine. Each agent consumes CPU, memory, and API rate limits. Start with 2-3 parallel agents and scale up based on your hardware and API quotas.
• Not reviewing agent output before merging. Parallel agents can produce conflicting architectural decisions. Review each worktree before merging to maintain code consistency.
• Assigning overlapping tasks to different agents. If two agents modify the same files, merging becomes complex. Assign tasks that touch different parts of the codebase.
• Failing to review community discussions and changelogs before upgrading. Breaking changes in major versions can disrupt existing workflows. Pin versions in production and test upgrades in staging first.