A Busy Bee is a strategic framework that orchestrates the best Claude Code tools into a seamless, token-efficient workflow. It automatically selects the cheapest tool for every job, spawns subagents to keep context clean, and learns from every session.
| Pain Point | Without A Busy Bee | With A Busy Bee |
|---|---|---|
| Token burnout | 10,000-15,000 per module | ~5,000 per module (60-70% savings) |
| Rate limits | Hit walls, wait forever | Smart retry + queuing |
| Context overflow | AI forgets mid-task | Auto-snapshot & fresh sessions |
A Busy Bee orchestrates three strategic tool groups, always picking the cheapest one that gets the job done:
| Tool | Role | Token Cost | When to Use |
|---|---|---|---|
| GSD | Planning, checklists, docs, tracking | ~200-400 | Every organizational task |
| Everything-Claude-Code | Coding, review (verdict and fix plan), security, rules | ~800-2,000 (via subagent) | Main development work |
| Superpowers | TDD, complex debugging, failure recovery | ~1,200-1,500 | Only when truly needed |
| Mem | Persistent memory across sessions | ~0 (file-based) | Lessons learned, decisions |
Result: Save 60-70% of tokens while delivering better code quality.
# 1. Clone the hive
git clone https://github.com/Olympusxvn/a_busy_bee.git
cd a_busy_bee
# 2. Open Claude Code
claude .
# 3. Deploy the bee
"Hey Busy Bee, plan a new module following the workflow." or cd "your_project"| Tool | Command |
|---|---|
| GSD | npx get-shit-done-cc@latest (run in terminal, choose Claude Code) |
| Everything | Visit everything-claude-code and follow instructions |
| Superpowers | Inside Claude Code: /plugin install superpowers@claude-plugins-official |
Even without installing the separate tools, you can use this repo as a configuration template -- just clone and open in Claude Code.
a_busy_bee/
├── .claude/
│ ├── settings.json # Token & context config
│ ├── agents/ # Planner (Opus) + Code Reviewer (Sonnet)
│ ├── common/ # 10 shared rules (security, testing, git, etc.)
│ ├── typescript/ # 5 TS-specific rules
│ ├── subagents/ # Planner, Coder, Debugger
│ └── skills/ # gsd-workflow, token-saver, mem
├── .github/
│ └── ISSUE_TEMPLATE/ # Bug report, feature request, question
├── docs/
│ ├── strategy.md # The 3-tool strategy explained
│ └── mem/ # Persistent memory store
├── templates/
│ ├── plan-template.md # Copy-paste for any module
│ └── spec-template.md # 5-10 line spec template
├── CONTRIBUTING.md # How to contribute
└── README.md # You are here
GSD: plan --> save to file
Everything (subagent): code according to spec
Everything (review): review code with code-reviewer agent
Superpowers (if error): debug with TDD
Mem: save lessons learned
A Busy Bee ships with curated rules from everything-claude-code:
Rules (.claude/common/ and .claude/typescript/):
coding-style, security, testing, performance, patterns, git-workflow, code-review, hooks, agents, development-workflow
Agents (.claude/agents/):
- Planner -- creates comprehensive implementation plans with phases, risk assessment, and testing strategies
- Code Reviewer -- confidence-based review covering security, code quality, performance, and best practices
| Benefit | What It Means For You |
|---|---|
| Save $ on tokens | 60-70% token reduction = lower API costs |
| Ship faster | Subagents work in parallel, not sequential |
| No more context crashes | Auto-snapshot before overflow |
| Learn once, apply forever | Persistent memory across sessions |
| Zero lock-in | Just templates + configs -- take them anywhere |
We welcome all forms of contribution -- from typo fixes to new subagents. See CONTRIBUTING.md for details.
- Report bugs: Open an issue
- Feature requests: Request a feature
- Share stories: How did A Busy Bee save your tokens?
First-time contributors welcome. Start with our good first issue label.
MIT -- Busy but free.
