Emdash: Parallel AI Coding Agent Desktop
Run multiple coding agents in parallel, each in its own Git worktree, and review & merge in one place.
Evidence shows: README claims local-first, data stored in local SQLite, telemetry optional and disableable, but no code-level verification. Deductions: no implementation details for least privilege, user confirmation, sensitive data handling; dependency security only lists build dependencies, no security audit; external effects (e.g., SSH connections, Git operations) lack explicit user confirmation mechanisms; rollback only mentions Git worktree, no app-level rollback; source attribution only mentions General Action, Inc., but publisher unverified.
Evidence shows: multiple database migration tests indicate internal consistency; but dependency availability unverified, failure messages not shown in source. Deductions: dependency availability relies solely on package manager, no redundancy or mirrors; failure messages not present in provided files.
Evidence shows: README describes multiple use cases (local, remote, multi-agent), supports multiple agents and platforms; capability boundaries mentioned in docs but not detailed; trigger precision not explicit; environment fit covers macOS, Windows, Linux. Deductions: trigger precision not evident in source; capability boundaries only asserted in docs, no code verification.
Evidence shows: README well-structured with installation, usage, privacy sections; install instructions detailed; license Apache-2.0; but naming stability, examples/FAQ, known limitations, versioning/changelog, maintenance responsibility not fully evident. Deductions: naming stability not evident in source; examples and FAQ only link to external docs; known limitations not listed; changelog not provided; maintenance responsibility only mentions contributing guide, no explicit maintainers.
Evidence shows: output usability via diff review, PR creation features; marginal value in parallel agents and integrations; cost-benefit not quantified. Deductions: cost-benefit lacks pricing or resource consumption info.
Evidence shows: README claims partially traceable to doc links, but no code-level verification; cross-source corroboration insufficient; facts and inferences not clearly separated. Deductions: claims mostly based on docs, no independent verification; facts and inferences mixed.
- Publisher identity unverified, proceed with caution.
- Dependency security not audited, supply chain risk.
- Remote SSH and Git operations may have external effects, ensure user confirmation mechanisms.
- Telemetry disableable but default state unclear.
What does this agent do, and when should you use it?
Emdash is a desktop application designed to run AI coding agents in parallel, replacing the need to juggle terminals. It leverages Git worktrees to give each task an isolated working directory and branch, letting you explore multiple fixes or features at once, review diffs, and merge what works. It works with local projects and remote machines over SSH, and supports the CLI agents you already use, such as Claude Code, Codex, OpenCode, and Amp. You can send issues and tickets from Linear, GitHub, Jira, GitLab, Asana, and other platforms into an agent, then review diffs, create PRs, inspect CI checks, and merge from a single interface. The app is local-first: state is stored in a local SQLite database, and Emdash does not send your code or chats to its own servers. It is available for macOS, Windows, and Linux under the Apache-2.0 license.
Emdash provides a GUI to manage the lifecycle of parallel coding agents. It automatically detects installed provider CLIs, including Claude Code, Codex, Cursor, OpenCode, and more, and creates a separate Git worktree and branch for each task to ensure isolation. You can import issues from Linear, GitHub, Jira, GitLab, etc., into tasks, then view each agent's diffs, create pull requests, inspect CI status, and merge from one dashboard. Remote projects are accessed via SSH/SFTP, supporting SSH agent, key, or password authentication. App state is persisted in a local SQLite database, and telemetry is optional, disableable via the TELEMETRY_ENABLED=false environment variable or in Settings.
- A full-stack developer wants to try two different bug fixes simultaneously without manually switching Git branches or terminals.
- A tech lead wants to assign multiple Linear issues to different coding agents in parallel and review the resulting PRs from one place.
- A remote team member maintains a codebase on a cloud VM and wants to run parallel agents directly on the remote machine via SSH without syncing locally.
- An indie developer using Claude Code wants to compare responses from Codex and OpenCode for the same task to pick the best implementation.
- An open-source maintainer needs to evaluate multiple feature implementations quickly by running an agent per feature in parallel and reviewing their diffs.
What are this agent's strengths and limitations?
- Runs multiple agents in parallel with each task isolated in a Git worktree, avoiding branch conflicts and boosting exploration efficiency.
- Integrates with major coding agents (Claude Code, Codex, OpenCode, etc.)—reuses tools you already know without learning a new agent.
- Directly imports issues from Linear, GitHub, Jira, etc., slotting into your existing project management workflow.
- Unified management for local or SSH remote projects: review diffs, PRs, CI, and merge in one place.
- Local-first architecture keeps data in a local SQLite database; code and chats are not sent to Emdash servers, enhancing privacy.
- As a desktop app, it requires installing a client on your OS—no headless/server usage.
- Depends on pre-installed agent CLIs (e.g., Claude Code, Codex) for actual AI inference; you must configure each provider's API credentials separately.
- Running multiple agents in parallel can be resource-intensive and may slow down low-spec machines.
- Remote functionality relies on SSH/SFTP, requiring extra network and authentication setup that may be blocked by firewalls or policies.
How do you install or deploy this agent?
Emdash offers platform-specific installers: for macOS, brew install --cask emdash or download the .dmg; for Windows, download the .msi or portable .exe; for Linux, use the AppImage or .deb package. All builds are available from the latest GitHub release.
How do you use this agent?
After launching Emdash, it automatically detects installed agent CLIs like Claude Code or Codex. Create a task, choose which agent to run, and optionally import an issue from Linear or GitHub. Each task runs in its own Git worktree; you can review diffs, create PRs, inspect CI, and merge from the interface. For remote projects, add an SSH/SFTP connection with key, password, or SSH agent auth. To disable telemetry, set TELEMETRY_ENABLED=false.