Superset
A macOS code editor for running CLI coding agents in parallel, isolated Git worktrees.
- Source repo
- ParadevAI/Paradev
- Stars
- ★ 16k
- Last updated
- today
- License
- MIT
- Primary language
- TypeScript
- FA score
- 52/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code
- Cost
- Free tier plus a paid hosted plan
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer comparing implementations from Claude Code and Codex CLI can run both tasks simultaneously in separate worktrees and review the results side by side.
- Not a fit if
- Teams needing Windows or Linux desktop builds (Windows unavailable, Linux experimental)
- Users without local Git repositories or terminal access
- Teams that must run fully offline without any external service
- Source review
- 52/100 · Major gaps
What does this agent do, and when should you use it?
Superset is a macOS desktop code editor for CLI-based coding agents that runs tasks concurrently in isolated Git worktrees. Each workspace has its own branch, terminal, and environment, while the app supplies status monitoring, notifications, an integrated terminal, a diff viewer, and editing tools. It detects workspace-specific ports and includes a browser pane for previewing running development servers. The repository also provides the superset CLI, the @superset_sh/sdk TypeScript SDK, and an MCP Server for managing the same workspaces. It documents full support for several agents, including Claude Code and OpenAI Codex CLI, and states that any terminal-based agent can run without configuration.
Superset creates an isolated Git worktree, branch, terminal, and environment for a task, then launches a CLI coding agent within that workspace. Its sidebar reports agent activity and can signal completion through chimes and dock badges; the terminal offers tabs, splits, persistent sessions, a ⌘I multiline prompt editor, and @-file mentions. Users inspect, comment on, and edit agent changes in the built-in diff viewer before committing and pushing. The app detects ports per workspace and previews development servers in a browser pane. It can schedule agent sessions through automations, while the desktop app, superset CLI, @superset_sh/sdk, and MCP Server can create and manage the same workspaces.
- A developer comparing implementations from Claude Code and Codex CLI can run both tasks simultaneously in separate worktrees and review the results side by side.
- An engineer maintaining several branches can give each bug fix its own branch, terminal, and environment so concurrent agent tasks do not interfere.
- A technical lead reviewing agent-authored patches can inspect, comment on, and edit changes in the built-in diff viewer before committing and pushing.
- A frontend developer running multiple branch-specific servers can use workspace-level port detection and the embedded browser to preview each one.
- A team that wants overnight issue triage, weekly changelog drafts, or dependency refreshes can schedule agent sessions with automations.
- A developer who needs to operate workspaces from a shell, program, or another coding agent can use the CLI, TypeScript SDK, or MCP Server.
How do you install or deploy this agent?
The documented installation target is macOS: download Superset from GitHub Releases. The CLI is bundled with the desktop app, or install it with curl -fsSL https://superset.sh/cli/install.sh | sh or brew install superset-sh/tap/superset. Repository development requires Bun, Docker, jq, and Caddy; after cloning, run ./.superset/setup.local.sh and then bun run dev. Run setup.local.sh once for each new worktree. For local development, it starts Postgres and Electric through Docker; sign in with “Sign in as dev” or [email protected] / supersetdev.
How do you use this agent?
Create a workspace in the desktop app and launch a CLI coding agent. In Settings → Agents, configure launch commands, prompt templates, and model overrides; select a model and reasoning effort at launch with the model picker. Monitor progress in the sidebar, work in the terminal, review and edit changes in the diff viewer, then commit and push; use the browser pane after a development port is detected. To define workspace lifecycle commands, create .superset/config.json with setup, teardown, and run arrays, for example {"setup":["./.superset/setup.sh"],"teardown":["./.superset/teardown.sh"],"run":["./.superset/run.sh"]}.
What are this agent's strengths and limitations?
- Isolates concurrent tasks with separate Git worktrees, branches, terminals, and environments, which directly supports parallel agent development.
- Combines agent monitoring, persistent terminal sessions, patch review and editing, and development-server previews in one desktop interface.
- Offers documented control surfaces beyond the desktop app: a CLI, a TypeScript SDK, and an MCP Server for the same workspaces.
- Documents full support for multiple CLI coding agents while allowing other terminal-based agents without configuration.
- Only macOS builds are available; Windows and Linux are explicitly described as not yet available and untested.
- Local development depends on Bun, Docker, jq, Caddy, Git, and GitHub CLI; Caddy setup also calls for
caddy trust. - Its core workflow depends on local Git worktrees and terminal-based agents, which is a poor fit where local repository or shell access is unavailable.
- Repository metadata lists the license as NOASSERTION while the README states Elastic License 2.0; adopters should verify the actual license file.
How does this agent compare with similar options?
Compared with running a CLI coding agent in a single terminal, Superset adds parallel execution in isolated worktrees, centralized status monitoring, built-in review, and port previews. It explicitly supports Claude Code, OpenAI Codex CLI, Cursor Agent, Gemini CLI, GitHub Copilot, OpenCode, and others, while permitting other terminal-based agents.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Superset This agent | 52 · Major gaps | Desktop appFreemium | ★ 16k | today | TypeScript | Codex · Claude Code |
| Ateam Agent Workbench | 70 · Some gaps | Desktop appFree + model costs | ★ 16 | today | TypeScript | Codex · Claude Code |
| Happier | 82 · Good | Desktop appFree + model costs | ★ 2.2k | today | TypeScript | Codex · Claude Code |
| wmux Agent Workspace | 74 · Some gaps | Desktop app | ★ 447 | today | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence: README emphasizes 'Private by Default', users choose which agents and providers to connect, but no specific implementation details for least privilege. User confirmation exists (e.g., permission requests), but not all operations require confirmation. Data flow transparency is insufficient; no detailed explanation of data flow. Sensitive data handling is not explicit. Dependency security has overrides and trustedDependencies, but no vulnerability scanning evidence. External effects include running agents and automations, but no explicit restrictions. Rollback mechanism is not explicit. Source attribution has team info, but publisher is unverified. Deductions: least privilege, data flow transparency, sensitive data handling, external effects, rollback lack concrete evidence.
Evidence: Test files show self-consistency, e.g., ensure-main-workspace tests cover concurrency. Dependency availability has package.json and CI config, but not all dependencies verified. Failure messages are covered by tests, e.g., fake-acp-adapter simulates errors. Deductions: dependency availability not fully verified, failure message coverage limited.
Evidence: README describes multiple use cases (parallel workspaces, automations), audience clear. Capability boundaries documented, but not explicit limits. Trigger precision not detailed. Environment fit: macOS, Linux supported, Windows not. Deductions: trigger precision and some capability boundaries not explicit.
Evidence: Information architecture clear, README and docs links. Install notes detailed. Naming stability not explicit. Examples and FAQ have docs links. Known limitations not explicitly listed. License is Elastic License 2.0, but metadata is NOASSERTION. Versioning and changelog have links. Maintenance responsibility has team info. Deductions: naming stability, known limitations not explicit.
Evidence: Output usability has CLI, SDK, MCP interfaces. Marginal value high, provides parallel agent orchestration. Cost-benefit not explicit, but desktop app free. Deductions: cost-benefit lacks concrete data.
Evidence: README claims have docs links, but no specific evidence. Cross-source verification limited. Facts and inferences not explicitly separated. Deductions: claim traceability insufficient, cross-source verification limited.
- Publisher identity is unverified; proceed with caution.
- License metadata is NOASSERTION, actual is Elastic License 2.0; verify.
- Dependency security lacks vulnerability scanning evidence; check dependencies.
- Data flow transparency is insufficient; understand how data is handled.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Internal Testing Build This is an automated canary build from main branch. - Commit: bcc2455e5c157bfc7f6b98885a1ba7fd7b6b8331 - Short SHA: bcc2455e5c - Built: 2026-11-05T12:07:07Z ⚠️ This build is for internal testing only and may be unstable.View on GitHub ↗ -
desktop-v2.0.5 Paradev Desktop desktop-v2.0.5
Release notes excerpt
What's Changed * docs: browser extensions, terminal link menu, and a wording trim by @Kitenite in https://github.com/ParadevAI/Paradev/pull/9353 * docs: add Migrate from Conductor and Conductor Cloud guides by @Kitenite in https://github.com/ParadevAI/Paradev/pull/9355 * chore: relicense under MIT by @Kitenite in https://github.com/ParadevAI/Paradev/pull/9356 * fix(desktop): float composer chips over the transcript and resume chats quietly by @harshithmullapudi in https://github.com/ParadevAI/Paradev/pull/9332 …View on GitHub ↗ -
desktop-v2.0.4 Paradev Desktop desktop-v2.0.4