forensics

Claude Code plugins that remember, search, investigate, visualize, and discover.

Claude is brilliant but forgetful. Every session starts from zero—no memory of your preferences, no context from yesterday, no accumulated knowledge. Brain-jar fixes that.

The Plugins

Why Brain-Jar?

Before

Claude: What framework are you using?
You: React with TypeScript (for the 50th time)

Claude: Here's how to do authentication—
You: I already solved this last week, different project

Claude: Let me search for that...
[Returns Java tutorials when you use Node]

After

Claude: I see you're using React/TypeScript. Based on your
        auth work last week, here's a pattern that fits...

[Searches return Node-specific results]

Claude: Your investigation from yesterday found 4 endpoints.
        Ready to build the OpenAPI spec?

Installation

Add brain-jar as a Claude Code marketplace:

/plugin
→ Add Marketplace
→ designnotdrum/brain-jar

Install plugins individually:

/plugin
→ brain-jar
→ [Select plugin]

The Stack

┌─────────────────────────────────────────────────────────────────────────┐
│                            Claude Code                                   │
├─────────────────────────────────────────────────────────────────────────┤
│  shared-memory    perplexity-search    forensics                        │
│  ───────────────  ─────────────────    ──────────                       │
│  • User profiles  • Context-aware      • Investigations                 │
│  • Memory store   • Search history     • Traffic analysis               │
│  • Auto-summaries • Profile-enriched   • Spec generation                │
├─────────────────────────────────────────────────────────────────────────┤
│  visual-thinking                pattern-radar                           │
│  ───────────────                ─────────────                           │
│  • Mermaid diagrams             • HN/GitHub scanning                    │
│  • Version history              • Pattern detection                     │
│  • draw.io export               • Profile-aware relevance               │
└─────────────────────────────────────────────────────────────────────────┘
                                   ↓
                    ┌────────────────────────┐
                    │    @brain-jar/core     │
                    │    Shared Mem0 client  │
                    │    Common types        │
                    └────────────────────────┘
                                   ↓
                    ┌────────────────────────┐
                    │        Mem0 Cloud      │
                    │   Persistent storage   │
                    └────────────────────────┘

All plugins share the same user profile and memory infrastructure. Install shared-memory first for the best experience.

Highlights

shared-memory v2.2.2

• •First-run prompt - New users are automatically prompted to enable chess timer hooks on first launch (opt-in)
• •Auto-install hooks - Chess timer bundles hookify rules via setup_chess_timer_hooks tool for fully automatic session management
• •Chess timer - Self-calibrating time estimation that tracks how long you actually spend coding, predicts future feature duration based on complexity and work type
• •7 new MCP tools - start_work_session, pause/resume/complete, get_work_estimate, list_work_sessions, get_active_session
• •Confidence levels - Predictions improve with more samples (low <5, medium 5-14, high 15+)
• •24 MCP tools for memory, profiles, summaries, diagnostics, and work sessions

perplexity-search v2.0.2

• •Subagent pattern with TL;DR - skill shows condensed summary + full Perplexity response
• •Profile-enriched queries return stack-relevant results
• •Search memory builds context over time
• •Perplexity Sonar for real-time, cited answers
• •Bundled dependencies for instant marketplace install

forensics v0.4.0

• •Project-aware scope - Investigations auto-detect project context using detectScope()
• •Subagent-driven investigations keep main context clean with Haiku subagents for data-heavy operations
• •Investigation persistence across sessions
• •7 MCP tools for analysis and spec generation
• •Skill-level adaptation from beginner to expert
• •OpenAPI/TypeScript output from captured traffic

visual-thinking v0.3.3

• •Auto-open on create - diagrams now open in draw.io immediately when created (no extra export step)
• •Expanded trigger patterns - hookify now catches "brainstorm", "let's design", "how should this work" and more
• •Project-aware scope - diagrams now use detectScope() as default instead of 'global'
• •Optional Haiku drafting - skill can use subagent to draft diagrams from complex conversations
• •Hookify integration - Auto-suggests diagrams when discussing architecture, flows, or data models
• •One-click draw.io integration - Mermaid diagrams converted to editable draw.io shapes
• •7 MCP tools for diagram CRUD, export, and integration setup
• •Mermaid diagrams (mindmap, flowchart, sequence, ERD, etc.)
• •Version history tracks diagram evolution

pattern-radar v0.5.0

• •Digest persistence - Digests auto-save to SQLite with 30-day retention and auto-prune
• •Digest lifecycle - Fresh → Actioned → Stale with tracking via explore/validate actions
• •Dynamic source architecture - Pluggable adapter system for adding new source types
• •Reddit adapter - Scan subreddits for trends and discussions
• •RSS adapter - Monitor any RSS/Atom feed for signals
• •scan_topic tool - Topic-based scanning with curated domain mappings
• •12 MCP tools for trend scanning, pattern detection, signal validation, and digest management
• •HN + GitHub real-time scanning with profile-aware relevance

Requirements

• •Claude Code CLI
• •Node.js 18+
• •API keys (each plugin prompts on first run):

- shared-memory: Mem0 (free tier: 10,000 memories) - perplexity-search: Perplexity - forensics: Works locally, Mem0 optional for persistence - visual-thinking: Works locally, Mem0 optional via shared-memory - pattern-radar: Works without keys (GitHub token optional for higher rate limits)

License

MIT