Add project files, examples, and documentation

- Added package.json for npm/bun compatibility
- Added tsconfig.json for TypeScript configuration
- Added MIT LICENSE
- Added CHANGELOG.md for version tracking
- Added 4 example files demonstrating usage:
  - 01-basic-compaction.ts - Context compaction basics
  - 02-claude-code-integration.ts - Claude Code integration
  - 03-openclaw-integration.ts - OpenClaw integration
  - 04-state-machine-pipeline.ts - Deterministic pipeline

All extracted source files are included in the repository.

examples/02-claude-code-integration.ts

@@ -0,0 +1,100 @@
+/**
+ * Example: Claude Code Integration
+ * 
+ * Demonstrates how to use the Claude Code integration
+ * for context-aware AI applications.
+ */
+
+import { ClaudeCodeIntegration } from '../agent-system/integrations/claude-code';
+
+async function claudeCodeExample() {
+  // Initialize Claude Code integration
+  const claude = new ClaudeCodeIntegration({
+    maxContextTokens: 200000,     // Claude's 200k context window
+    reserveTokens: 40000,         // Reserve for response
+    compactionStrategy: 'hybrid',
+    autoCompact: true,            // Auto-compact when threshold reached
+    compactionThreshold: 0.75,    // Compact at 75% capacity
+    enableSubagents: true,        // Enable subagent spawning
+    maxSubagents: 6,
+    persistentMemory: true,
+    memoryStorePath: '.claude-code/memory'
+  });
+
+  console.log('Session ID:', claude.getSessionId());
+
+  // Add user message
+  claude.addMessage({
+    role: 'user',
+    content: 'Analyze this codebase and identify potential security issues.',
+    metadata: {
+      priority: 1,
+      fileReferences: ['src/auth.ts', 'src/middleware.ts']
+    }
+  });
+
+  // Check token stats
+  const stats = claude.getTokenStats();
+  console.log(`Token usage: ${stats.percentage}% (${stats.used}/${stats.total})`);
+
+  // Get context for Claude API
+  const { messages, systemPrompt } = claude.getContextForAPI();
+  console.log(`Prepared ${messages.length} messages for API`);
+  if (systemPrompt) {
+    console.log('System prompt includes context summary');
+  }
+
+  // Spawn a researcher subagent
+  console.log('\n--- Spawning Researcher Subagent ---');
+  const researcherResult = await claude.spawnSubagent({
+    type: 'researcher',
+    prompt: 'Research best practices for JWT authentication in TypeScript',
+    priority: 'high'
+  });
+  
+  console.log('Researcher result:', {
+    success: researcherResult.success,
+    tokens: researcherResult.tokens,
+    duration: `${researcherResult.duration}ms`
+  });
+
+  // Spawn multiple subagents in parallel (OpenClaw pattern: 4×3)
+  console.log('\n--- Parallel Subagent Execution ---');
+  const parallelResults = await claude.executeParallelSubagents([
+    { type: 'explorer', prompt: 'Find all authentication endpoints' },
+    { type: 'explorer', prompt: 'Find all database queries' },
+    { type: 'reviewer', prompt: 'Review input validation logic' }
+  ]);
+
+  console.log(`Completed ${parallelResults.length} subagent tasks`);
+
+  // Use persistent memory
+  await claude.remember('projectContext', {
+    name: 'MyApp',
+    type: 'REST API',
+    securityScan: 'in-progress'
+  });
+
+  const savedContext = await claude.recall('projectContext');
+  console.log('Recalled context:', savedContext);
+
+  // Save context for later restoration
+  await claude.saveContext('security-audit-milestone');
+  console.log('Context saved for future sessions');
+
+  // Get session info
+  const session = claude.getSessionInfo();
+  console.log('\nSession Info:', {
+    id: session.id,
+    messageCount: session.messageCount,
+    tokenUsage: session.tokenUsage,
+    status: session.status
+  });
+
+  // Get registered tools
+  const tools = claude.getTools();
+  console.log(`\nRegistered tools: ${tools.map(t => t.name).join(', ')}`);
+}
+
+// Run the example
+claudeCodeExample().catch(console.error);