Tyran plugin
A Claude Code plugin as a manager: interviews you, cuts the job into tickets and hands them to subagents on a model tier per piece
- 79 ★ on the repository
- Last commit
Broad access
Why this level
- Plugin hooks decide which tool calls pass, and agents edit files and branches and may push within the chosen autonomy class
- The `curl | sh` installer puts the plugin and a secrets scanner on your machine
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/tyranDetects 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.
- Plugin hooks decide which tool calls pass, and agents edit files and branches and may push within the chosen autonomy class
- The `curl | sh` installer puts the plugin and a secrets scanner on your machine
This is third-party code. Review the repository files before installing.
What it does
Tyran is a Claude Code plugin that turns one long chat into managed work. You describe the job and the conductor asks clarifying questions in batches of at most four, each with a recommendation, turns the plan into tickets on a board and hands each piece to a separate subagent: scout, implementer, reviewer or retro. A subagent gets fresh context and a model tier matched to difficulty, and security review always runs on the strongest tier. Decisions, spawns, reports and merges are written to a journal committed to your repository. A hook decides before a tool runs, so the rules do not rest on a request in the prompt. A loopback dashboard shows agent activity, pending questions and token spend per model.
Who it is for. Developers running large tasks in Claude Code who want to see the plan, token spend and reviewer verdict, and teams where several people work in one repository.
Good fit when
- The task spans many files and one chat fills with plan and logs
- You want a reviewer to check a patch it did not write itself
- An idea arrives mid-run and must become a ticket without getting lost in chat history
- Model spend grew and you want simple repository reading sent to a cheap tier
- A team switches between branches and must see a shared decision journal
Not a fit when
- The task is a single edit: the interview and tickets only slow things down
- Claude Code is older than 2.1: the plugin hooks and agents are unsupported
- Running a `curl | sh` installer and installing a secrets scanner is forbidden for you
- You need work with no journal and state written into the repository
Example request
Run Tyran setup in this repository, then interview me and split the payment module refactor into ticketsLimitations
It needs Claude Code 2.1 or newer. The client must be restarted after install: hooks and agents load at session start. By default Tyran refuses to read `.env`, write outside the repository and push beyond the chosen autonomy class; `boundaries: {preset: open}` in `.tyran/config.yaml` relaxes those five boundaries. Secret scanning at commit and push cannot be turned off. State and journal live in `.tyran/` inside the repository. To stop every agent write a line to `.tyran/STOP`.
How to disable. Remove the plugin through `/plugin` and the `.tyran/` folder from the repository if you do not need the journal. To halt work instantly write a line to `.tyran/STOP`.
FAQ
How do I install Tyran?
With one command: `curl -fsSL https://raw.githubusercontent.com/jjanczur/tyran/main/install.sh | sh`. Or manually in Claude Code: `/plugin marketplace add jjanczur/tyran`, `/plugin install tyran@tyran` and `/tyran:hello`, then restart.
What does `/tyran:setup` do?
Once per repository it scans the project, writes `.tyran/` and installs the `/tyran` command. It infers stack, validation commands and autonomy class from how the repo is actually worked and asks only about what it could not establish.
How do I open the board?
Setup turns autostart on: each session starts the dashboard and prints the URL. It is bound to loopback and refreshes every 30 seconds.
How do I stop all agents at once?
Run `echo "wrong branch — hold everything" > .tyran/STOP` in the project root. The session stays alive while work holds.
Which subagents are included?
`tyran:scout`, `tyran:implementer`, `tyran:reviewer` and `tyran:retro`. The reviewer can never bless a patch it wrote and security review runs on the strongest tier.
Discussion
No questions or reviews yet.
Related
PluginsMulti-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
Claude Code plugin with 60 subagents and 62 team presets, plus a browser designer for planning agent teams visually.
Tonone installs 140 specialist agents in 15 groups and 573 skills into Claude Code, from engineering to legal and finance.
A Claude Code plugin that draws a 3D office where your session and subagents work at desks from real transcripts, zero config.
Claude Code supervises while Codex, opencode and pi work on your quota: delegated edits, independent review, model discussions
claude-mods adds 112 skills, 3 agents and 3 commands to Claude Code: stacks, agent operation, defense, session continuity.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Tyran (this card) | no | 79 | |
| swarm-orchestrator | no | 114 | a swarm CLI; Tyran lives inside Claude Code through hooks, tickets and a token spend board |
| CLI Agent Orchestrator | no | 1,548 | orchestration of several CLI agents; here one client and a split by model tier |
README in short
The README explains the manager-instead-of-one-chat idea, cost and quality as two levers, the journal in the repository and self-improvement through retrospectives. It then covers install, commands, the stop file, the dashboard and docs links.