Superset is a source-available agentic IDE and CLI for running many AI coding agents at the same time. Each task gets its own git worktree with a separate branch, terminal, and environment, so agents never step on each other's work. It's agent-agnostic: bring Claude Code, Codex, Cursor Agent, Gemini CLI, Copilot, OpenCode, or any CLI tool with your own subscription. You review every change in a built-in diff viewer and ship through normal branches and PRs. Built for developers tired of waiting on one agent at a time.

• Parallel agent orchestration: Launch dozens of coding agents at once, each isolated in its own git worktree. You stop waiting for one agent to finish before starting the next task.
• Agent-agnostic by design: Works with Claude Code, OpenAI Codex, Cursor Agent, Gemini CLI, GitHub Copilot, Amp, OpenCode, and any CLI-based agent. You use your existing subscriptions and API keys with no markup added.
• Git worktree isolation: Every workspace runs on its own branch and directory while sharing git history, which avoids merge conflicts and saves disk space compared to multiple clones.
• Central monitoring dashboard: Track every running agent from one place. Status indicators, completion chimes, dock badges, and push notifications tell you exactly which agent needs input.
• Built-in diff viewer and editor: Review each agent's changes inline before merging, with a chat pane supporting tool approvals and plan review.
• One-click editor handoff: Open any worktree in VS Code, Cursor, JetBrains, Xcode, Sublime Text, or a plain terminal without losing context.
• CLI, SDK, and MCP server: Drive workspaces from the command line, a TypeScript SDK, or as an MCP server so other tools can spin up agents programmatically.
• Local-first privacy: Repos, worktrees, terminal output, and agent sessions stay on your machine. Cloud sync covers only account and organization metadata.
• GitHub, Linear, and Slack integrations: Create workspaces straight from an issue or a Slack message, with persistent terminal sessions and automated port management.

• Free for solo developers: The Free plan costs $0, includes local workspaces and the desktop app, and requires no credit card. You can also use your existing agent subscriptions and API keys.
• Works with your existing AI agents: Use Claude Code, Codex, Amp, Gemini, OpenCode, Cursor Agent, and other CLI agents without being locked to a single provider.
• Parallel development workflows: Run features, bug fixes, refactors, and reviews in separate workspaces instead of switching between tasks in a single branch.
• Local-first privacy: Repositories, worktrees, terminal output, and agent sessions stay on your machine by default. Cloud sync covers account and organization metadata.
• Automated worktree management: Superset creates isolated Git workspaces for your tasks and handles worktree cleanup when you delete completed workspaces.

• macOS-first platform support: Official builds target macOS. Linux arrived via AppImage more recently and Windows remains unsupported in the README, so Windows users must build from source themselves.
• Source-available, not open source: The Elastic License 2.0 restricts commercial redistribution, which matters if you planned to fork or resell it. Some third-party sites incorrectly list it as Apache 2.0.
• More agents mean more code to review: Running many agents in parallel can shift the bottleneck from waiting for code to reviewing and integrating multiple sets of changes.

• Parallel feature development for solo founders: Assign one agent to a new feature, another to a bug fix, and a third to test coverage. Each runs on its own branch, so you wake up to three separate PRs instead of one sequential change.
• Matching agents to task strengths: Different agents excel at different work. Put Claude Code on a complex refactor while Codex generates tests at the same time, then compare results side by side in the diff viewer.
• Batch delegation for tech leads: Break a sprint backlog into non-overlapping tasks and dispatch them all at once. You keep review control while the mechanical work happens concurrently.
• Issue-to-workspace automation: Connect Linear or Slack and spin up a fully configured agent workspace directly from a ticket or a message, skipping the manual setup of branches and environments.
• Docs and maintenance cleanup: Send low-risk work like documentation updates, dependency bumps, and lint fixes to background agents while you focus on the product code that needs your attention.
• Comparing agent output on the same task: Run two or three agents against the same prompt in separate worktrees, then pick the best implementation and discard the rest without any merge cleanup.