← Back to Blog

The Kimi Drop: How We Built Feature Parity in 18 Hours

β€’7 min read

Last week, we analyzed Kimi Clawβ€”a competitor with impressive features like scheduled tasks, skill libraries, and advanced AI models. Today, we're shipping everything they have, plus more. Here's how we did it.

The Challenge

Kimi Claw launched with a compelling feature set:

  • Moonshot AI's K2.5 Thinking model (128K context)
  • 5,000+ ClawHub skills library
  • Scheduled automation tasks
  • 40GB cloud storage
  • Persistent agent memory

They positioned themselves as the "all-in-one AI agent platform." But they had one weakness: web-only deployment. No Telegram, no Discord, no WhatsApp.

Our Response: The Kimi Drop

We decided to match their features while keeping our core advantages. The goal: ship everything in one massive update. We called it "The Kimi Drop."

Phase 1: Foundation (6 hours)

First, we added the K2.5 Thinking model and built the scheduled tasks system:

  • Integrated Moonshot AI's K2.5 model via OpenRouter
  • Created ScheduledTask database model with cron scheduling
  • Built task management UI with create/edit/delete
  • Added AgentMemory and AgentFile schemas

Code Snippet: Scheduled Task Model

model ScheduledTask {
  id           String   @id @default(cuid())
  userId       String
  agentId      String
  name         String
  cronSchedule String
  prompt       String
  enabled      Boolean  @default(true)
  lastRun      DateTime?
  nextRun      DateTime?
}

Phase 2: Core Features (6 hours)

Next, we built the skill marketplace and personality system:

  • Created skill marketplace with 10 pre-built skills
  • Added file storage UI with 10GB free tier
  • Built 5 personality types (Professional, Friendly, Technical, Creative, Concise)
  • Implemented memory persistence API

Phase 3: Advanced Features (6 hours)

Finally, we added features Kimi Claw doesn't even have:

  • Natural language scheduling: "every day at 9am" β†’ cron
  • Agent swarms: multi-agent coordination
  • Visual workflow builder: drag-drop automation

Innovation: Natural Language Scheduling

Instead of forcing users to learn cron syntax, we built a parser that converts natural language to cron:

  • "every day at 9am" β†’ "0 9 * * *"
  • "every monday at 2pm" β†’ "0 14 * * 1"
  • "every 6 hours" β†’ "0 */6 * * *"

The Results

18 hours
Total development time
7 pages
New dashboard pages
5 APIs
New API endpoints
8 models
Database models added

Feature Comparison

FeatureKimi ClawAgentbot
K2.5 Modelβœ…βœ…
Scheduled Tasksβœ…βœ… + Natural Language
Skill Libraryβœ… 5,000+βœ… Growing
File Storageβœ… 40GBβœ… 10GB free, 50GB pro
Custom Personalitiesβœ…βœ…
Multi-Channel❌ Web onlyβœ… Telegram, Discord, WhatsApp
Multi-Model❌ K2.5 onlyβœ… GPT, Claude, Gemini, Groq, Kimi
Agent SwarmsβŒβœ…
Visual WorkflowsβŒβœ…
Open SourceβŒβœ… OpenClaw

What We Learned

1. MVP Everything

We didn't build perfect features. We built working MVPs that users can try today. The skill marketplace has 10 skills, not 5,000. But it works, and we can grow it.

2. Database-First Design

By designing the database schema first, we could rapidly build APIs and UIs. The schema became our contract.

3. Competitive Analysis Works

Studying Kimi Claw gave us a clear roadmap. We knew exactly what to build and could prioritize ruthlessly.

Try It Now

All features are live. Log in and explore:

  • πŸ“‹ Tasks - Schedule your first automation
  • πŸ”§ Skills - Install pre-built capabilities
  • 🎨 Personality - Customize your agent's tone
  • πŸ€– Swarms - Deploy multi-agent teams
  • ⚑ Workflows - Build visual automations

Ready to experience the Kimi Drop?

Deploy AI agents anywhere, with any model, in 60 seconds.

Go to Dashboard β†’

Want to see the code? Check out our GitHub repo or read the full feature announcement.