TL;DR
Context Buckets are persistent containers of knowledge injected into AI conversations. Instead of explaining your project from scratch every time, you build a knowledge base that makes Claude progressively smarter about YOUR specific codebase.
The Context Bucket Methodology
Every Context Bucket follows a simple formula:
01Project Container
Top-level folder representing a business domain. Think “Marketing”, “Sales Pipeline”, “Core Architecture”, or “Legal”.
02Goals (Grit Monitors)
Persistent objectives that guide AI behavior across 100+ prompts. Not checkboxes — habits and standards to maintain.
03Accumulated Context
Markdown files containing domain knowledge, SOPs, patterns, and history. The AI learns from this over time.
Why Context Buckets Work
Traditional AI assistants are stateless. Every conversation starts fresh. You explain the same architecture decisions, the same code patterns, the same business rules — repeatedly.
Without Context Buckets
- -Re-explain project every session
- -Generic code suggestions
- -Repeat same mistakes
- -No learning over time
With Context Buckets
- +Context injected automatically
- +Project-specific suggestions
- +Gotchas already documented
- +Compounding intelligence
Key Features
Markdown File Injection
Add any markdown file to your context. Each file is token-counted so you know exactly how much context you're using.
Hierarchical Organization
3-level outline structure. Organize by domain, then topic, then detail. Drag-and-drop to reorder.
Token Usage Tracking
See token counts per document and total. Optimize your context to stay within model limits while maximizing information density.
Clipboard Export
One click to copy all selected context to clipboard, ready for injection into any Claude conversation.
Persistent Storage
All context documents stored in local SQLite database. Your knowledge base persists across sessions and restarts.
Example: Full Business Context
Here's how a startup CEO might organize their Context Buckets:
Company Context/
├── Marketing/
│ ├── brand-guidelines.md (5,200 tokens)
│ ├── campaign-templates.md (3,100 tokens)
│ └── competitor-analysis.md (8,400 tokens)
├── Sales/
│ ├── pricing-model.md (2,800 tokens)
│ ├── objection-handling.md (4,200 tokens)
│ └── pipeline-stages.md (1,900 tokens)
├── Finance/
│ ├── budget-2025.md (6,700 tokens)
│ ├── cash-flow-model.md (5,100 tokens)
│ └── investor-updates.md (3,400 tokens)
├── Operations/
│ ├── architecture.md (12,000 tokens)
│ ├── deployment-sops.md (4,800 tokens)
│ └── incident-playbook.md (3,600 tokens)
└── Legal/
├── terms-of-service.md (8,900 tokens)
├── privacy-policy.md (7,200 tokens)
└── contractor-agreements.md (5,500 tokens)Total: ~83,000 tokens of business context, organized and ready for injection.
Token Economics
Understanding token usage helps you optimize your context for different scenarios:
| Context Level | Typical Size | Use Case |
|---|---|---|
| Minimal | 5,000 tokens | Quick questions, simple fixes |
| Standard | 20,000 tokens | Feature development |
| Comprehensive | 50,000 tokens | Strategic planning, architecture |
| Maximum | 100,000+ tokens | Full business context |
Best Practices
Start Small, Grow Organically
Begin with your most critical context (architecture, patterns). Add more as you discover what's useful.
Document Gotchas Immediately
When you discover a bug or edge case, add it to context right away. Never solve the same problem twice.
Add Descriptions
Each document can have a description. Use it to remind yourself (and the AI) why this context matters.
Prune Regularly
Outdated context is worse than no context. Review and update your documents when the project evolves.
Next Steps
Now that you understand Context Buckets, explore how to persist goals across sessions.