ELVES

What ELVES is

ELVES is a macOS desktop app for running AI agents in isolated Git worktrees. Describe a task, review the resulting changes as a diff, then merge the branch or discard its worktree. ELVES works with Claude Code and OpenAI Codex, and multiple sessions can run at the same time in separate tabbed workspaces called floors. Its persistent memory uses SQLite and FTS5 search, and can automatically supply an agent with relevant saved memories before a task. The app also includes MCP server management and can import MCP servers from Claude. Each task can include up to 10 attached files, with a combined size limit of 500KB. The maker says projects, memory, and sessions are stored locally, without cloud services, accounts, or telemetry. ELVES is free and released under the MIT License. The macOS app requires macOS 13 or later, Node.js 18 or later, and either the Claude Code CLI or Codex CLI. It is not notarized; the maker provides instructions for opening it despite Gatekeeper warnings.

Who it suits

ELVES suits developers who want to coordinate Claude Code or OpenAI Codex work across isolated Git worktrees and review changes before merging. It may also suit people who want project data and sessions stored locally and need concurrent agent sessions, persistent memory, or MCP server management.

Limits

ELVES is free. Each task can include up to 10 attached files totaling 500KB, and the app requires macOS 13 or later, Node.js 18 or later, and the Claude Code CLI or Codex CLI.

Getting started

1. Visit the ELVES website.
2. Use a Mac running macOS 13 or later and install Node.js 18 or later.
3. Install the Claude Code CLI or Codex CLI.
4. Get the free ELVES app and follow the maker’s instructions to open it despite Gatekeeper warnings.
5. Describe a task, review its diff, then merge the branch or discard the worktree.

Verdict

Choose ELVES if you use Claude Code or OpenAI Codex on macOS and want to manage agent tasks in separate worktrees, inspect their diffs, and decide whether to merge. Its local storage and persistent memory may also suit developers who want those capabilities in one desktop app. Look elsewhere if you need a notarized macOS app or do not meet its system and CLI requirements.