SupaConductor plugin
A plugin that turns one command into a loop: spec, plan, parallel agent work, review by evaluators and fixes
- 383 ★ on the repository
- Last commit
- AGPL-3.0
Broad access
Why this level
- In agentic mode the system writes code, runs commands and makes decisions without asking the user
- Several agents work in parallel and sharply raise token spend
- The virtual board gives a rationale but does not replace your decision about a production deploy or deleting data
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/supaconductorDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- In agentic mode the system writes code, runs commands and makes decisions without asking the user
- Several agents work in parallel and sharply raise token spend
- The virtual board gives a rationale but does not replace your decision about a production deploy or deleting data
This is third-party code. Review the repository files before installing.
What it does
SupaConductor installs into Claude Code as a plugin and runs a chain called the Evaluate-Loop. You state a goal with `/orchestrator-supaconductor:go` and get a spec, a task plan with dependencies, execution and review. Independent tasks run in parallel across agents. Four evaluators check the interface, code quality, integrations and business logic. When a check fails, a fix step runs: up to five fix cycles and up to three plan revisions. The kit includes 36 commands, 39 skills, 15 agents, four lead engineers (architecture, product, tech and QA) and a board of directors with five roles. The board discusses major decisions and votes with written rationale: `board-meeting` gives a full deliberation and `board-review` a quick assessment. It joins automatically on sensitive tracks such as a production deploy or a breaking API migration. A `conductor` folder with work tracks appears in the project, and the mode is chosen in `conductor/config.json`: agentic, where the system decides on its own, or human-in-the-loop, where it stops at key decisions. The plugin ships together with Superpowers version 4.3.0.
Who it is for. For developers who hand Claude Code larger tasks across three or more files and want an automatic plan, review and fixes instead of manual re-prompting.
Good fit when
- You are adding a multi-part feature, such as Google sign-in with token handling, and want a spec and plan before code
- The plan's tasks are independent and worth running at the same time across several agents
- You want finished code checked by evaluators for interface, quality and integrations, not only by tests
- Before a major deploy or migration decision you want a review from several roles and a written rationale
- An interrupted session must resume from the point where work on a track stopped
Not a fit when
- A change in a single file: plain Claude Code is enough for it
- The token budget is tight: the system uses roughly three to five times more API calls than manual work in one conversation
- You need full control of every step: in agentic mode the system decides on its own and asks nothing
- The AGPL-3.0 license conflicts with your organization's policy
Example request
/orchestrator-supaconductor:go Add Stripe payments with webhooks, split the work into independent tasks and run the evaluators on the result.Limitations
The reference platform is Claude Code. For Codex, Cursor, Antigravity, Windsurf and GitHub Copilot you build the adapters yourself: clone the repository and run `python3 scripts/build-platform-adapters.py`, and ready packages appear in `dist/platforms`. These clients differ in capability, and a detailed table sits in docs/platform-support.md.
The author estimates three to five times more calls than manual work: the plugin uses Opus for planning and evaluation and Sonnet for execution. Skills load on demand, and agents work in separate conversations so they do not fill the main context.
The license is AGPL-3.0, and the Superpowers skills that ship with the plugin are MIT-licensed. Work state lives in the `conductor` folder of your project. Availability of Anthropic models from Russia was not checked when this card was prepared.
How to disable. In Claude Code run `/plugin uninstall orchestrator-supaconductor@ibrahim-plugins`. Delete the `conductor` folder in the project separately if you no longer need the tracks and plan.
FAQ
How do I install SupaConductor in Claude Code?
Run `/plugin marketplace add Ibrahim-3d/orchestrator-supaconductor`, then `/plugin install orchestrator-supaconductor@ibrahim-plugins`. Run `/orchestrator-supaconductor:setup` once: the command studies the project, creates the `conductor` folder and a first set of tracks.
What does the go command do in SupaConductor?
It takes a goal in plain text and runs the whole loop: writes a spec, builds a plan with dependencies, executes the tasks, checks them with evaluators and fixes findings. If a session breaks off, running it again with no arguments resumes the work.
How many tokens does SupaConductor use?
The author gives a rule of thumb of three to five times more API calls than manual work in one conversation. You cut the cost by skipping board meetings for small features, running `implement` with your own spec and choosing the human-in-the-loop mode.
How does SupaConductor differ from Superpowers?
Superpowers gives an agent a development process through a set of skills, while SupaConductor adds an automatic loop on top with parallel agents, evaluators and a board of directors. Superpowers version 4.3.0 ships with it.
Does SupaConductor work in Codex and Cursor?
Yes, through adapters generated by the `build-platform-adapters.py` script. Packages for Codex, Cursor, Antigravity, Windsurf and GitHub Copilot appear in `dist/platforms`, and releases attach ready ZIP files.
Discussion
No questions or reviews yet.
Related
PluginsClaude Code plugin with 60 subagents and 62 team presets, plus a browser designer for planning agent teams visually.
An orchestration plugin for Claude Code, Codex, Gemini CLI and Qwen Code: 39 specialists, parallel subagents, persistent sessions and standalone audits
Multi-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
A Claude Code plugin that hands coding tasks to Codex agents, monitors them and verifies the results independently
Orchestration layer for Claude Code, Cursor and Codex: a board of 495 expert personas gives a verdict with named dissent, plus hundreds of skills
CLI Agent Orchestrator
CLI Agent Orchestrator (CAO)
A CLI that coordinates several coding agents: a supervisor delegates tasks to specialists in parallel or in sequence across isolated tmux sessions
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| SupaConductor (this card) | no | 383 | |
| Superpowers | no | 316,679 | a skill set with a development process, with no automatic review loop or board of directors |
| Conductor | no | 3,791 | Google's plugin for spec-driven development where context, spec and plan come before code |
README in short
The README explains the difference between working with an agent by hand and with one command, lists what the kit contains and shows installation for Claude Code and other clients. Separate sections cover the evaluate loop, parallel execution, the two modes, the board of directors and a command reference. The FAQ touches on cost, context use and when the system is overkill.