A collection of specialized agents that drastically reduce conversation context usage through intelligent summarization and background processing. Achieve 95%+ context savings while maintaining full functionality.
This plugin is not yet in any themed marketplace. To install it, you'll need to add it from GitHub directly.
Choose your preferred installation method below
A marketplace is a collection of plugins. Every plugin gets an auto-generated marketplace JSON for individual installation, plus inclusion in category and themed collections. Add a marketplace once (step 1), then install any plugin from it (step 2).
One-time setup for access to all plugins
When to use: If you plan to install multiple plugins now or later
Step 1: Add the marketplace (one-time)
/plugin marketplace add https://claudepluginhub.com/marketplaces/all.jsonRun this once to access all plugins
Step 2: Install this plugin
/plugin install context-preserving-agents@allUse this plugin's auto-generated marketplace JSON for individual installation
When to use: If you only want to try this specific plugin
Step 1: Add this plugin's marketplace
/plugin marketplace add https://claudepluginhub.com/marketplaces/plugins/context-preserving-agents.jsonStep 2: Install the plugin
/plugin install context-preserving-agents@context-preserving-agentsA collection of specialized agents for Claude Code that drastically reduce conversation context usage through intelligent summarization and background processing. Achieve 95%+ context savings while maintaining full functionality.
This plugin provides 6 specialized agents that handle complex tasks in the background, returning only concise summaries to keep your conversation context clean and efficient. Perfect for large codebases, complex analysis, and long development sessions.
Deep, comprehensive analysis without context bloat
research/[timestamp]-[topic].mdWhen to use: "analyze", "investigate", "understand", "research"
Structured analytical reasoning for complex problems
When to use: Complex decisions, multi-factor analysis, architectural choices
Preserves conversation context across task switches
.claude/session-notes.mdWhen to use: Long conversations, switching tasks, recalling previous work
Analyzes large files without cluttering context
When to use: Large log files, test output analysis, build logs
Complete documentation management
When to use: Any documentation task (README, API docs, guides)
Runs bash commands with verbose output, returns summaries
When to use: dotnet test, npm build, git log, large output commands
cp -r context-preserving-agents ~/.claude/plugins/
cd ~/.claude/plugins/
git clone https://github.com/yourusername/context-preserving-agents.git
/plugin install context-preserving-agents
The plugin activates automatically. Agents are invoked based on task patterns:
User: "Investigate the authentication system architecture"
→ Researcher Agent automatically handles this
User: "Should we use Redux or Context API? Consider performance and complexity"
→ Thinker Agent provides structured analysis
User: "Update the API documentation with the new endpoints"
→ Bookworm Agent manages documentation
For automatic agent routing, add the included CLAUDE.md template to your project:
cp context-preserving-agents/CLAUDE.md.template your-project/CLAUDE.md
This enables the decision tree that automatically routes tasks to the appropriate agent.
You can also invoke agents explicitly using the Task tool:
Use the researcher-agent to analyze the caching strategy
Use the thinker-agent to evaluate database options
Use the bookworm-agent to update README.md
┌─ Need to run bash/shell command? → bishbash-agent
├─ "analyze", "understand", "investigate"? → researcher-agent
├─ Documentation read/write/update? → bookworm-agent
├─ File > 1000 lines? → file-warden-agent
├─ Complex reasoning/planning? → thinker-agent
├─ Multi-step workflow? → general-purpose agent
└─ Need conversation history? → historian-agent
Based on real-world usage:
| Agent | Avg Context Saved | Typical Use Case | Processing Time |
|---|---|---|---|
| Researcher | 98.5% | Architecture analysis | 30-90s |
| Thinker | 89.2% | Decision analysis | 20-60s |
| Historian | 90.0% | Conversation summary | 15-45s |
| File Warden | 95.0% | Log file analysis | 10-30s |
| Bookworm | 92.0% | Documentation updates | 15-60s |
| BishBash | 97.0% | Command execution | 5-120s |
User: "I'm worried about security in our authentication system"
→ Researcher Agent investigates:
- Authentication flow security
- Token handling
- Session management
- Vulnerability patterns
→ Saves 15,000 character detailed report
→ Returns 5 critical findings (500 characters)
→ Context saved: 96.7%
User: "Run the full test suite and let me know the results"
→ BishBash Agent executes:
- Runs: npm test (2,500 lines of output)
- Identifies: 3 failing tests
- Extracts: Error messages and stack traces
→ Returns: Summary with failure details (200 characters)
→ Context saved: 98.4%
User: "Should we use microservices or monolith? Consider our team size and scaling needs"
→ Thinker Agent analyzes:
- Team capacity (5 developers)
- Current complexity
- Scaling requirements
- Maintenance overhead
→ Uses 12 thinking blocks
→ Returns: Structured recommendation (800 characters)
→ Context saved: 89.2%
/plugins listagents/ directory.claude/logs/Contributions welcome! Please:
MIT License - see LICENSE file for details
See CHANGELOG.md for version history
Developed for the Claude Code community to solve the context bloat problem that emerges in complex, long-running development sessions.
Start saving context today! Install and watch your conversation efficiency skyrocket.
1.0.0