(\(\
     (-.-)
     o_(")(")

Multi-Tier AI Orchestration for Cost-Optimized Development

Get Started · Felix · Commands · MCP Server

🚀 Cut AI Costs by 90%

Stop overpaying for AI. Krolik Felix automatically routes tasks to the optimal AI model tier based on complexity:

• Simple tasks → Free/Cheap models (formatting, refactoring, simple fixes)
• Medium tasks → Mid-tier models (feature implementation, debugging)
• Complex tasks → Premium models (architecture design, complex algorithms)

Result: 90% cost reduction compared to always using premium models.

Installation

npm i -g @anatolykoptev/krolik-cli

krolik felix — Multi-Tier AI Routing

Intelligent model orchestration that saves money without sacrificing quality.

Core Features

🎯 Automatic Tier Selection

Analyzes task complexity and routes to optimal tier:

• Free tier (Llama, DeepSeek) — 60% of tasks
• Cheap tier (Haiku, Flash) — 25% of tasks
• Mid tier (Sonnet, Pro) — 10% of tasks
• Premium tier (Opus, O1) — 5% of critical tasks

⚡ Cascade Fallback

If a cheaper model fails, automatically escalates to next tier:

Free → Cheap → Mid → Premium

Ensures reliability while maintaining cost efficiency.

📊 History Learning

SQLite database tracks success/failure patterns:

• Routes similar tasks to models that performed well historically
• Improves accuracy over time
• Learns project-specific patterns

💰 Cost Estimation

krolik felix estimate --prd PRD.json
# Optimistic: $2.50
# Expected: $4.80
# Pessimistic: $8.20

Usage

# Start autonomous execution from PRD
krolik felix start --prd PRD.json

# Get routing plan (dry-run)
krolik felix plan --prd PRD.json

# Check status
krolik felix status

# View routing statistics
krolik felix stats

Example PRD

{
  "name": "Add user authentication",
  "tasks": [
    {
      "id": "create-user-model",
      "description": "Create Prisma User model with email/password",
      "complexity": "low",
      "files_affected": ["prisma/schema.prisma"]
    },
    {
      "id": "implement-jwt-auth",
      "description": "Implement JWT-based authentication with refresh tokens",
      "complexity": "high",
      "files_affected": ["lib/auth.ts", "lib/jwt.ts"]
    }
  ]
}

Felix analyzes each task and routes:

• create-user-model → Cheap tier (simple schema addition)
• implement-jwt-auth → Premium tier (security-critical, complex logic)

Other Commands

Krolik also provides developer productivity tools:

krolik context — Project Context in Seconds

One command replaces 10+ manual searches.

krolik context --feature auth      # Everything about auth feature
krolik context --issue 42          # Context from GitHub issue
krolik context --quick             # Compact mode (~3500 tokens)

What it collects: git state, database schema, API routes, project structure, types, past decisions.

krolik mem — Persistent Memory

AI forgets everything between sessions. Krolik remembers.

krolik mem save --type decision --title "Use tRPC" --description "Type-safe API"
krolik mem search --query "authentication"
krolik mem recent --limit 5

Memory types: decisions, patterns, bugfixes, observations, features.

krolik audit — Find Quality Issues

Analyzes codebase and creates a prioritized report.

krolik audit                       # All issues
krolik audit --mode release        # Security + type-safety
krolik audit --mode pre-commit     # Quick check before commit

krolik fix — Auto-Fix Issues

krolik fix --dry-run        # Preview changes
krolik fix                  # Apply safe fixes

What it fixes: console.log, debugger, any types, magic numbers, and more.

krolik refactor — Find Duplicates

krolik refactor              # Find duplicate functions
krolik refactor --deep       # + type duplicates
krolik refactor --apply      # Apply migrations

krolik agent — Specialized AI Agents

krolik agent --list                           # Available agents
krolik agent security-auditor                 # Run specific agent
krolik agent --orchestrate --task "review"    # Multi-agent mode

Other Utilities

| Command | What it does | |---------|--------------| | krolik status | Quick health check: git, types, lint | | krolik schema | Database schema as structured docs | | krolik routes | API routes as structured docs | | krolik review | Code review for current changes | | krolik docs | Library documentation cache | | krolik modules | Query lib modules in your codebase |

MCP Server

Native integration with Claude Code via Model Context Protocol.

claude mcp add krolik -- npx @anatolykoptev/krolik-cli mcp

Configuration

Works out of the box. Customize if needed:

// krolik.config.ts
import { defineConfig } from '@anatolykoptev/krolik-cli';

export default defineConfig({
  name: 'my-project',
  paths: {
    web: 'apps/web',
    api: 'packages/api',
  },
});

Requirements

• Node.js >= 20.0.0
• TypeScript >= 5.0.0 (peer dependency)

Native modules require build tools:

• macOS: xcode-select --install
• Linux: apt install build-essential python3
• Windows: npm install -g windows-build-tools

Why "Felix"?

Felix (Latin for "lucky" or "successful") represents the tool's mission: making AI development cost-effective and successful for everyone. The rabbit mascot embodies speed and agility in navigating complex AI model choices.

Contributing

See CONTRIBUTING.md for development setup.

License

FSL-1.1-Apache-2.0 © Anatoly Koptev

Free for internal use, education, and research. Converts to Apache 2.0 on December 26, 2027.

Changelog

v0.18.0 (2026-01-18)

• Felix multi-tier routing — primary feature repositioning
  • Intelligent model selection across 4 tiers (Free, Cheap, Mid, Premium)
  • Cascade fallback for reliability
  • SQLite-based history learning
  • Cost estimation and routing statistics
  • 90% cost reduction compared to always using premium models
• Updated positioning — Multi-tier AI orchestration as primary value proposition
• Website launch — krolik.tools with product documentation

v0.17.0 (2026-01-14)

• Toma semantic clone detection — fast token-based duplicate detection
  • 65x faster than ML-based approach (84ms for 1600+ functions)
  • Phase 1 (O(n), hash-based) enabled by default
  • Phase 2 (O(n²), fuzzy) in deep mode only
  • 6 similarity metrics: Jaccard, Dice, Jaro-Winkler, Cosine, LCS

v0.16.0

• Context-aware console detection (skip debug files)
• Smart magic number naming based on context

v0.15.1 (2026-01-13)

• Cleaned up README, added changelog
• Major cleanup of refactor component
• Optimized MCP tools with direct imports
• Consolidated duplicate utility functions

v0.15.0

• Memory graph: link decisions with caused, supersedes, implements
• Semantic search with local AI model
• Added krolik docs for library documentation caching
• Added krolik modules for querying lib exports

v0.14.0

• Added krolik agent with specialized agent categories
• Added krolik refactor for duplicate detection
• Multi-agent orchestration mode