> ## Documentation Index
> Fetch the complete documentation index at: https://docs.chucky.cloud/llms.txt
> Use this file to discover all available pages before exploring further.
# Build AI agents with Claude
> Claude as a Service - Deploy AI agents with managed infrastructure
# Chucky SDK
Chucky provides **Claude as a Service** - a fully managed infrastructure for deploying AI agents powered by Claude. Build conversational AI applications without managing servers, handling authentication complexity, or worrying about scaling.
## The Problem
Building AI features into your product sounds simple: call the API, get a response, show it to users. But once you ship, reality hits:
| Challenge | What You Need to Build | Time |
| ------------- | ----------------------------------------------------------------------- | --------- |
| **Billing** | Usage tracking, budget limits, Stripe metering, usage-based billing | 2-4 weeks |
| **Execution** | Sandboxed environments, file system isolation, process management | 4-8 weeks |
| **Tools** | Tool execution routing, browser-to-server communication, error handling | 2-4 weeks |
| **Sessions** | Persistence, context management, message history, resume logic | 1-2 weeks |
That's **9-18 weeks** of infrastructure work before you can ship your AI feature.
## The Chucky Solution
Chucky handles all of this with one SDK:
```typescript theme={null}
// Billing, execution, tools, sessions - all handled.
const client = new ChuckyClient({ token });
const result = await client.prompt('Deploy my code to production');
```
**Built-in per-user billing:** JWT tokens carry budgets. Enforcement happens at the edge.
```typescript theme={null}
const token = await createToken({
userId: 'user-123',
budget: createBudget({
aiDollars: 5.00, // User can spend up to $5
computeHours: 1, // And 1 hour of compute
window: 'day', // Resets daily
}),
});
```
When the budget runs out, requests stop. No surprise bills.
## Comparison
| Capability | Direct API | Chucky |
| ------------------- | ----------------- | ---------------- |
| Per-user billing | Build it yourself | Built-in |
| Budget enforcement | Build it yourself | Automatic |
| Code execution | Build it yourself | Managed sandbox |
| File operations | Not available | Full access |
| Shell commands | Not available | Full access |
| Browser tools | Build it yourself | Native support |
| Local execution | Build it yourself | Possession Mode |
| Session persistence | Build it yourself | Automatic |
| Multi-turn context | Build it yourself | Automatic |
| Git integration | Build it yourself | Built-in bundles |
## Who Uses Chucky
**Agencies Billing Clients:** Deploy once, bill each client automatically. JWT tokens carry per-client budgets. No custom metering code per project.
**SaaS Teams Adding AI:** Add AI features without building infrastructure. Set budgets per subscription tier. Ship in days, not months.
**Indie Developers:** One SDK, one deployment. Focus on your product, not on building AI infrastructure.
## Key Features
One SDK, same API for browser and Node.js. WebSocket-based real-time communication.
Set per-user AI and compute budgets. Track usage in real-time with automatic limits.
Define tools that Claude can call. Execute in browser or server context.
Use Claude, OpenAI, or 100+ OpenRouter models. Configure behavior with environment variables.
## Supported Models
| Provider | Models | Best For |
| -------------- | -------------------------- | -------------------------- |
| **Anthropic** | Claude Opus, Sonnet, Haiku | Default, best quality |
| **OpenAI** | GPT-5.2, GPT-5, o3, o4 | Alternative provider |
| **OpenRouter** | 100+ models | Cost optimization, variety |
Configure models per-session with [environment variables](/concepts/models#environment-variables) like `ANTHROPIC_MODEL` and `CLAUDE_CODE_SUBAGENT_MODEL`.
## Next Steps
Get up and running in 5 minutes
How Chucky works under the hood
Practical guides and recipes
Pricing and budget controls
## Popular Topics
Multi-provider & env vars
Create custom tools
Connect external services
Command-line interface
Background tasks
Multi-turn conversations