vibecosystem plugin
A kit of 138 agents, 296 skills and 74 hooks that turns Claude Code into a full software development team
- 540 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The installer writes agents, skills, rules and hooks into `~/.claude`, and hooks fire on every tool call
- Agent chains in the full profile run commands and change code without pausing at every step
- Self-learning stores rules from past errors and applies them in other projects
Install
- Works in
Manual install
npx vibecosystem initThe main route; the lean core profile is installed by default.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The installer writes agents, skills, rules and hooks into `~/.claude`, and hooks fire on every tool call
- Agent chains in the full profile run commands and change code without pausing at every step
- Self-learning stores rules from past errors and applies them in other projects
This is third-party code. Review the repository files before installing.
What it does
vibecosystem is a ready-made kit for Claude Code: 138 specialized agents (from frontend-dev to security-analyst), 296 skills, 74 TypeScript hooks and 20 behavior rules. In the full profile a phrase such as "add a new feature" starts more than twenty agents. They go through five phases: scouting with a plan, development, review, a QA loop, and a final summary with self-learning. Hooks watch tool calls and add task-specific context. Self-learning records repeated errors as patterns and over time turns them into permanent rules for every project. Profiles set the size of the kit: a lean one is installed by default, and the full one enables every component. After installation the vibeco command is available. It shows statistics and the component list, switches profiles, checks the installation (`vibeco doctor`) and starts a monitoring panel (`vibeco dashboard`). Codex gets a luna_worker agent, and Cursor and OpenCode have their own rules.
Who it is for. For developers who want team-style work from Claude Code with planning, review, tests and accumulated rules without manual setup.
Good fit when
- A large feature needs planning, development, review and verification in one process
- You want ready-made specialist agents, for example for GraphQL, Kubernetes, security and tests
- Repeated mistakes should become rules for every project
- Tokens are limited: the lean profiles load only the agents and skills you need
- You want to watch agent activity in the `vibeco dashboard` panel
Not a fit when
- You need a small edit in one file: the kit adds token overhead
- You need a minimal setup: the kit holds hundreds of components that take time to learn
- You need an unchanging kit: the authors often ship large updates
Example request
Add an endpoint that exports orders to CSV: make a plan, implement it, review it and check it with tests.Limitations
The core profile caps context at 4000 characters per event and 12000 per session, while full allows 8000 and 50000. The larger the profile, the more agents and hooks take part in a task and the higher the token spend.
Node.js 18 or newer is required. The main platform is Claude Code, and the repository carries separate config files for Codex, Cursor and OpenCode. The numbers in the repository description and in the docs differ slightly (295 and 296 skills, 73 and 74 hooks), and `vibeco stats` shows the actual set.
How to disable. The `vibeco profile core` command trims the kit to the lean profile. To remove the kit entirely, delete the installed agents, skills, rules and hooks from `~/.claude` together with their entries in `settings.json`. The installer keeps ownership manifests that show which files it created.
FAQ
How do I install vibecosystem?
The main route is `npx vibecosystem init`. Alternatives are `curl -fsSL https://raw.githubusercontent.com/vibeeval/vibecosystem/main/install-remote.sh | bash` or a `git clone` of the repository followed by `./install.sh --profile core`.
How do the profiles differ?
`core` installs 12 agents and 32 skills for bounded implementation and verification. `memory` adds memory and learning, `orchestration` adds explicit agent chains, `frontend`, `backend`, `fullstack` and `devops` are domain sets, and `full` has every component. Switch with `vibeco profile <name>`.
How does self-learning work?
An error is recorded as a pattern. Once a pattern repeats five times it enters the context; five repeats across two projects make it shared by all of them, and after ten repeats it becomes a permanent rule. This belongs to the `memory` profile.
Where does the kit store its data?
All data stays on your machine in `~/.claude/`, and self-learned rules go to `~/.claude/rules/`. The kit makes no network requests and has no telemetry or cloud sync.
How many hooks run in the lean profile?
There are 74 hook sources, but the core profile registers only 6 commands, while full enables them all. `vibeco effective-config` shows the current profile, model and budgets.
Discussion
No questions or reviews yet.
Related
PluginsWorkflows for Claude Code: /do, /omo, /bmad-pilot and more commands, with Codex, Gemini or OpenCode writing the code through codeagent-wrapper
Multi-agent orchestration for Claude Code: teams, roles, autopilot and task routing across skills and subagents
A pack of agents, skills and hooks with a router: the /do command picks a specialist and drives the work through checks
A desktop app where Claude Code, Codex, Gemini and other agents work as a team of specialists in a shared project
claude-code-sub-agents
Claude Code Subagents Collection
A collection of specialized subagents for Claude Code: frontend, backend, data, infrastructure, testing, security
A large kit for Claude Code: subagents, skills, commands, hooks, rules, templates and MCP configs in one plugin marketplace
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| vibecosystem (this entry) | no | 540 | |
| oh-my-claudecode | no | 40,377 | multi-agent orchestration inside Claude Code with teams, roles, autopilot and task routing |
| Citadel | no | 927 | a single /do entry point, project memory across sessions, safety hooks and cost telemetry |
| Claude Code Toolkit | no | 2,786 | a collection of tools and configurations for Claude Code |
README in short
The docs keep a history of large releases: from v2.0 with a security scanner and a monitoring panel to v3.4, where the lean `core` profile became the default. They are translated into more than ten languages, including Russian (`docs/README_RU.md`). Upgrade details between versions are collected in `UPGRADING.md`.