Membria VSCode Extension

Membria decision intelligence integration for VSCode. Brings decision context, plan validation, and skill recommendations directly into your editor.

Features

📋 Decision Management

• Capture Decisions: Record decisions with alternatives and confidence levels
• Record Outcomes: Track decision results to build calibration data
• Decision Tree: Browse all decisions organized by status (Recent, Success, Failed, Pending)

🎯 Plan Validation

• Validate Plans: Get real-time feedback on plan steps
• Detection: Identifies conflicts with negative knowledge and antipatterns
• Warnings: Shows calibration gaps and debiasing suggestions

⭐ Skills & Recommendations

• List Skills: Browse domain-specific skills with quality scores
• Generate Skills: Create skills from outcomes and patterns
• Zones: See patterns categorized as Green (use confidently), Yellow (review carefully), Red (avoid)

📊 Team Calibration

• View Calibration: Track confidence vs success rates by domain
• Gaps: See where team is overconfident/underconfident
• Trends: Monitor improvement over time

🔍 Inline Features

• Hover Context: Hover over code to see decision context and warnings
• Decorations: Visual indicators for detected decision patterns and warnings
• Syntax Highlighting: Color-coded zones and pattern quality

Installation

Prerequisites

• VSCode 1.60 or later
• Node.js 14+
• Python 3.8+ (for Membria CLI)
• MCP Server running (see Claude Integration)

Steps

1. Clone the repository:

cd vscode-extension
npm install

2. Install and start the MCP server:

# In membria-cli directory
pip install -e .
python src/membria/start_mcp_server.py

3. Build the extension:

npm run compile

4. Press F5 to open the Debug session and test the extension

Package for Distribution

npm install -g @vscode/vsce
vsce package

This creates a .vsix file you can share or publish to the VSCode Marketplace.

Configuration

Add to .vscode/settings.json:

{
  "membria.serverHost": "localhost",
  "membria.serverPort": 6379,
  "membria.autoValidatePlans": true,
  "membria.showHoverContext": true,
  "membria.showInlineWarnings": true
}

Commands

Sidebar Views

📌 Decisions

Browse decisions organized by status and outcome type. Click to view details.

📊 Calibration

Team calibration metrics by domain showing:

• Success Rate: Historical success percentage
• Sample Size: Number of decisions observed
• Confidence Gap: Team calibration (overconfident/underconfident)
• Trend: Improving/Stable/Declining

⭐ Skills

Skills generated from outcomes and patterns. Shows:

• Skill ID and domain
• Quality Score: Evidence-based (success_rate × √sample_size / √n)
• Success Rate: Empirical success rate
• Zones: Green (>75%), Yellow (50-75%), Red (<50%)

📋 Plans

Team plans organized by status (Completed, In Progress, Pending). Shows:

• Plan ID and domain
• Step count and accuracy
• Last updated timestamp

How It Works

Decision Capture Flow

1. Run command: membria.captureDecision
2. Enter decision statement, alternatives, and confidence
3. Decision is sent to MCP server
4. Stored in FalkorDB with timestamp
5. Immediately appears in Decisions panel

Plan Validation Flow

1. Run command: membria.validatePlan
2. Enter plan steps (one per line)
3. MCP server checks against:
   • Negative Knowledge (known failures)
   • AntiPatterns (common mistakes)
   • Past Failures (team history)
   • Calibration (confidence gaps)
4. Returns warnings with severity levels
5. Can proceed if no high-severity issues

Skill Generation Flow

1. Sufficient outcomes collected in domain (≥30)
2. Patterns extracted from outcomes
3. Skill procedure generated with evidence
4. Quality score calculated: success_rate × (1 - 1/√sample_size)
5. Available via membria.generateSkill command

Architecture

Components

• membriaClient.ts: HTTP client to MCP server
• providers/: Tree data providers for sidebar views
• hoverProvider.ts: Hover context display
• decorationProvider.ts: Inline warnings and indicators
• extension.ts: Main activation and command registration

Data Flow

VSCode Extension → MCP Server (HTTP) → FalkorDB
                                    ↓
                    CalibrationUpdater, PatternExtractor, SkillGenerator

Troubleshooting

MCP Server Connection Failed

• Check server is running: python src/membria/start_mcp_server.py
• Verify port matches settings (default: 6379)
• Check firewall if running on different machine

No Skills Generated

• Need ≥30 outcomes per domain (sample size requirement)
• Run: membria skills generate <domain> via CLI
• Check calibration data exists for domain

Hover Context Not Showing

• Enable in settings: "membria.showHoverContext": true
• Check server connectivity
• File must be Python/TypeScript/JavaScript

Plan Validation Takes Too Long

• Server may be querying graph database
• First query is slower, subsequent queries are cached
• Adjust timeout in membriaClient.ts if needed

Contributing

Contributions welcome! Areas for enhancement:

• Additional hover information (related decisions, evidence)
• Decision timeline visualization
• Skill comparison charts
• Integration with source control (git)
• Webview panels improvements

License

MIT

Related

• Membria CLI - Command-line interface
• Claude Integration - Claude Code integration
• VSCode Tasks Integration - Task automation