Superset

repository·main·Indexed 11 days ago

https://github.com/superset-sh/superset

A specialized code editor designed to orchestrate and manage swarms of CLI-based AI coding agents using git worktrees to provide isolated, parallel development environments.

Tokens
815.1K
Snippets
2.9K
Records
4.6K
Agent score
93%

What's inside Superset

  1. Overview of discord-triage

    main

    The discord-triage package is a gateway bot that automatically ingests Discord support messages into Linear Triage.

    Core Workflow:

    • Every new top-level message in a watched text channel or new post in a watched forum channel is converted into a Linear issue labeled Source: Discord.
    • The bot opens a thread on the original Discord message containing a link to the newly created Linear issue.
    • Enhancements: If ANTHROPIC_API_KEY is provided, the bot uses Claude Sonnet to rewrite the ticket into a standard format (improving title, Context, Artifacts, and References) while preserving the original report as a quote. It also mirrors message attachments to Linear uploads to prevent link expiration.
    • Webhook Support: If LINEAR_WEBHOOK_SECRET is set, the bot listens at /linear-webhook to archive the corresponding Discord thread when an issue is closed in Linear.
  2. Overview of Superset

    main
    Superset is a code editor designed for AI agents. It orchestrates CLI-based coding agents (such as Claude Code or Codex) across isolated git worktrees. This allows developers to run multiple agents simultaneously in parallel workspaces without context switching, as each agent operates in its own branch, terminal, and environment. Superset provides built-in tools for monitoring agents, reviewing changes via a diff viewer, and managing workspaces through a CLI, SDK, or MCP.
  3. Overview of @superset/pty-daemon

    main

    The @superset/pty-daemon package provides a long-lived process that owns Pseudo-Terminals (PTYs) for the Superset v2 desktop terminal. Its primary purpose is to ensure that terminal sessions and shells remain active even when the host-service restarts or when the daemon binary itself is upgraded (via file descriptor handoff).

    Key characteristics:

    • Standalone: It does not import from other workspace packages; it only shares protocol types via @superset/pty-daemon/protocol.
    • Stateless Protocol: Every protocol call carries full context. The daemon does not track client state or maintain session tombstones.
    • Security: Authentication is handled at the OS level via Unix socket file modes (0600). The daemon trusts any client capable of opening the socket.
    • In-Memory Buffer: Terminal scrollback is kept in an in-memory 64KB ring buffer per session. It is not persisted to disk.
  4. Overview of Superset for AI Coding Agents

    main

    Superset is an engineering-grade workspace designed for running AI coding agents in parallel. Unlike general-purpose agent apps, Superset provides Git-native isolation by assigning every task its own Git worktree and branch. It is designed for developers who need agents to ship code through real diffs and PR-centric workflows.

    Key Features:

    • Git-native Isolation: Every task operates in its own worktree and branch.
    • Integrated Tooling: Includes persistent terminals, a diff/file editor, an in-app browser, and port management.
    • Agent Support: Runs Claude Code, Codex, OpenCode, Cursor, Copilot, Gemini, and any other CLI-based agent.
    • Programmability: Supports scheduled automations, a TypeScript SDK, an MCP server, and a Slack bot.

    Platform Support:

    • macOS (Current)
    • Windows and Linux (Coming soon)
  5. Overview of Superset as an Agent Workspace

    main

    Superset is a local-first desktop workspace designed for orchestrating AI coding agents (such as Claude Code, Codex, OpenCode, Cursor Agent, Copilot, Gemini CLI, and Mistral Vibe).

    Key features include:

    • Git Worktree Isolation: Automatically creates an isolated Git worktree for every task to ensure environment stability.
    • Persistent Sessions: Maintains persistent terminal sessions, diffs, and review states within each worktree-backed workspace.
    • Integrated Tooling: Includes a built-in diff/file editor, chat panel, in-app browser (for documentation and dev servers), port management, and MCP (Model Context Protocol) tooling.
    • Remote/Cloud Capabilities: Supports running workspaces across network devices via remote and cloud workspace features.
    • Automations: Supports scheduled agent sessions, a TypeScript SDK for composing workflows, an MCP server for agent-driven control, and a Slack bot for triggering and receiving run updates.

    Superset is source-available under the Elastic License 2.0 (ELv2).

  6. Overview of Superset as an Agent Orchestrator Alternative

    main

    Superset is a workspace for running fleets of AI coding agents in isolated worktrees. It is designed to provide a more complete development environment than Agent Orchestrator (AO) by including built-in tools for review and automation.

    Key features include:

    • Workspace Tools: Built-in diff/file editor, in-app browser, port management, and persistent daemon-backed terminals.
    • Extensibility & Scripting: A maintained CLI, an MCP (Model Context Protocol) server, a TypeScript SDK, and a Slack bot.
    • Automation: Scheduled automations (cron) and the ability to surface PR status, review threads, and CI checks directly in the workspace.
    • Collaboration: Remote and cloud workspaces for multi-device use and team-based plans.

    Platform Support: macOS (current); Windows and Linux (upcoming).
    License: Source-available (ELv2).

  7. Overview of Superset as a Mux alternative

    main

    Superset is an agent-agnostic orchestrator designed for parallel agentic development. Unlike Mux, which uses its own proprietary agent loop and bills via API tokens, Superset allows you to run any CLI agent (such as Claude Code, Codex, OpenCode, Cursor, Copilot, or Gemini) while preserving your existing configurations, skills, and subscription-based billing (e.g., Claude Max or ChatGPT plans).

    Key features include:

    • Automatic Git Worktrees: Each agent runs in an isolated worktree.
    • Persistent Environments: Includes persistent terminals, a diff/file editor, an in-app browser, and port management.
    • Orchestration: Fully scriptable via a CLI, MCP server, and TypeScript SDK.
    • Automations: Supports scheduled sessions.
    • Scalability: Supports remote and cloud workspaces that extend across devices.
  8. What is Superset?

    main

    Superset is a source-available platform (Elastic License 2.0) designed for running AI coding agents in parallel. It provides an environment where every task is assigned its own workspace, which is an isolated git worktree on its own branch. This allows multiple agents to work simultaneously without interfering with each other or your main branch.

    Superset is available in three forms:

    • A desktop IDE
    • A command-line interface (CLI)
    • An MCP server

    Key features include:

    • Worktree isolation: Automated management of git worktrees to prevent manual stashing or checkout conflicts.
    • Agent Agnostic: Supports Claude Code, Codex, Amp, Gemini, OpenCode, Cursor Agent, and more, as well as a built-in chat.
    • Integrated Tooling: Includes terminals, diff viewers, commit/PR management, port forwarding, and dev servers, all scoped to the specific workspace.
    • Local-first: Works offline and keeps code on your machine unless explicitly synced.
    • Automations: Supports scheduled agent sessions that output to real workspaces.
  9. Overview of Superset Agentic Development Environment

    main

    Superset is a local-first desktop workspace designed for running AI coding agents in parallel. It provides isolation for every task using automatic Git worktrees, ensuring that each agent session operates in its own branch and environment.

    Key features include:

    • Agent Support: Runs any CLI-based agent (e.g., Claude Code, Codex, OpenCode, Cursor Agent, Copilot, Gemini CLI, Mistral Vibe) inside persistent terminal sessions.
    • Integrated Tooling: Built-in diff/file editor, chat panel, in-app browser for documentation and dev servers, port management, and MCP (Model Context Protocol) tooling.
    • Workspaces: Supports local-first operation with the ability to move whole workspaces across network devices via remote and cloud workspaces.
    • Automation: Allows scheduling agent sessions through a TypeScript SDK, an MCP server, and a Slack bot for programmatic workflow composition.
  10. Compare Superset and Orca for agentic workflows

    main

    Superset and Orca are both desktop workspaces designed to run multiple AI coding agents in parallel using Git worktree isolation.

    Superset is a local-first workspace focused on workflow depth and team automation. It supports scheduled agent sessions, a TypeScript SDK, a Slack bot, and an MCP server. It is source-available under the Elastic License 2.0 (ELv2) and offers cloud/remote workspaces that move entire workspaces across network devices.

    Orca is an open-source (MIT) agent development environment (ADE) focused on solo-developer breadth. It features 'Design Mode' (sending UI elements/screenshots to agents), a 'fan-out' workflow (running one prompt across multiple agents to compare results), and mobile companion apps for iOS and Android. Its remote capability centers on SSH/remote worktrees on a VPS or other machines.

  11. Compare Superset and Agent Orchestrator (AO)

    main

    Superset and Agent Orchestrator (AO) are both desktop applications designed to run fleets of AI coding agents in parallel using isolated Git worktrees.

    Key Distinctions

    • Superset is a local-first agent workspace focused on a deep developer experience. It provides a full workspace including a diff/file editor, in-app browser, persistent terminals, port management, and MCP tooling. It is designed for extensibility via a CLI, TypeScript SDK, MCP server, and Slack bot, and supports scheduled automations. It also supports remote and cloud workspaces across network devices.
    • Agent Orchestrator (AO) is an agentic IDE focused on autonomous feedback loops. Its core feature is a 'reactions' system that automatically routes CI failures, review comments, and merge conflicts back to agents as configurable prompts. It is primarily local-only and lacks a public API or scheduling capabilities.

    Comparison Summary

    FeatureSupersetAgent Orchestrator (AO)
    Primary FocusFull workspace + automations/SDK/MCPAutomatic CI/review/conflict feedback loops
    Agent SupportAny CLI agent (Claude Code, Codex, etc.)23 specific harnesses (Claude Code, aider, etc.)
    IsolationAutomatic Git worktree per taskWorktree per session (or full clone)
    AutomationScheduled sessions, SDK, Slack bot, CLIEvent-driven reactions only
    Remote/TeamRemote/cloud workspaces, team/enterprise tiersLocal-only, no built-in auth for remote access
    LicenseSource-available (ELv2)Open source (Apache-2.0)