Emdash Lets You Run 20 Coding Agents at Once and Actually Keep Track of Them

The Macro: The Agent Swarm Needs a Control Room

I run coding agents every day. Sometimes three or four at a time. And the honest experience is chaos. Each one lives in its own terminal tab, its own window, its own little silo. I forget which one is doing what. I lose track of which branch got modified. I accidentally step on an agent’s work because I forgot it was still running.

This is the dirty secret of agentic coding in 2026: the agents themselves have gotten pretty good, but the workflow around them is still a mess. You are the orchestrator, the project manager, the merge conflict resolver, and the babysitter. All at once. All manually.

The market knows this is a problem. Cursor gives you one agent at a time with a nice UI. Claude Code runs beautifully but lives in the terminal. Codex operates in its own sandbox. Windsurf, Copilot, Augment, Cline. The list goes on. But none of them are really built for the use case of “I want five agents working on five different tasks in the same repo simultaneously, and I want to see what all of them are doing.”

That is the gap Emdash is going after. Not another coding agent. A dashboard for all your coding agents.

The timing makes sense. As agents get more capable, the bottleneck shifts from “can the agent write good code” to “can I manage all these agents without losing my mind.” It is a coordination problem, not an intelligence problem. And coordination problems need purpose-built tools.

The Micro: Git Worktrees, Kanban Boards, and 220K Downloads

Emdash is a desktop application. Open source, provider-agnostic, and built around one clever architectural decision: Git worktrees. Each agent you spin up gets its own isolated worktree, which means it can make changes without stepping on any other agent’s work. No branch conflicts. No accidental overwrites. Each agent operates in its own clean copy of the repo.

The UI is a Kanban board. Each card represents an agent task. You can see what every agent is doing, review its diffs inline, and approve or reject changes before they hit your main branch. It supports over 20 agents out of the box, including Codex, Cursor, Claude Code, Copilot, Cline, and a long tail of others. The CLI auto-detects what you have installed and makes them available.

Remote SSH is supported, which matters for teams working on distributed repos or running agents on beefy remote machines. MCP server integration is in there too.

The product already has over 220,000 downloads, which is a meaningful number for a developer tool that launched recently. The open source angle helps, obviously. But downloads at that scale suggest the pain point is real and widespread.

Arne Strickmann and Raban von Spiegel are the cofounders. Raban previously founded Soff (a YC S24 company) and did research at MIT CSAIL on software performance engineering. Arne comes from Langdock and Highlight AI. Both have been in the developer tooling space long enough to understand the workflow problems they are solving. The company came through Y Combinator’s W26 batch.

The pricing question is interesting for an open source product. Emdash is free to use, and the GitHub repo is public. The monetization path likely involves enterprise features, team coordination, or hosted infrastructure, but that is not fully visible yet.

What I find most compelling is the philosophical bet: agents are going to multiply, not consolidate. We are not heading toward one agent that does everything. We are heading toward specialized agents that each do one thing well, running in parallel. If that is true, then the management layer becomes essential infrastructure, not a nice-to-have.

The Verdict

Emdash is solving the right problem at the right time. The agent-per-task model is clearly where things are heading, and the Git worktree approach is genuinely smart. Isolation without duplication. Parallelism without chaos.

At 30 days, I want to know: does the Kanban view actually help, or do power users just ignore it and go back to the terminal? The diff viewer needs to be better than what you get in VS Code, or people will not use it.

At 60 days: how does it handle repos with hundreds of thousands of files? Worktrees are lightweight, but twenty of them in a monorepo could get heavy fast.

At 90 days: what does the team workflow look like? If five engineers are each running five agents, that is twenty-five parallel worktrees. The coordination challenge scales quadratically, and Emdash needs to prove it can handle that without becoming its own bottleneck.

I am cautiously optimistic. The download numbers suggest real pull. The open source model builds trust. And the core insight, that agents need management more than they need improvement, feels correct.