The Vercel for AI Agents

Expose Claude Code as a production-ready API

Cloud brain, local hands. Built-in per-user billing. Deploy in 60 seconds.

✓ Browser, server & host control✓ Per-user budget enforcement✓ Universal SDK (JS + Python)

app.tsx

import { ChuckyClient, getAssistantText } from '@chucky.cloud/sdk';

// Tools run in YOUR browser, Claude calls them from the sandbox
const showAlert = {
  name: 'show_alert', description: 'Show a browser alert',
  inputSchema: { type: 'object', properties: { message: { type: 'string' } } },
  handler: async ({ message }) => {
    alert(message);
    return { content: [{ type: 'text', text: 'Alert shown' }] };
  }
};

const session = client.createSession({
  mcpServers: [{ name: 'browser', version: '1.0.0', tools: [showAlert] }]
});
await session.send('Show an alert saying hello!');

Architecture

How it works

Your code calls our cloud. Claude reasons securely. Tools execute in your environment.

Your App

SDK or CLI

•Browser, Node.js, or Python
•JWT token with budget
•Same API everywhere

client.prompt("...")

prompt

prompt↓

Chucky Cloud

Secure sandbox

•Claude + your workspace
•MCP servers + tools
•Budget enforcement

Claude reasons here

tool_call

tool_call↓

Your Environment

Browser or host

•Tool handlers (mcpServers)
•Host tools (possession)
•Local execution

Actions happen here

Unique Capability

Possession Mode

Claude reasons in the cloud. Your machine executes. One command unlocks AI-powered automation for your entire workflow.

1. You run CLI

chucky prompt "..." --allow-possession

2. Claude reasons

Secure sandbox in Chucky cloud

3. Host tools execute

Actions run on YOUR machine

Available Host Tools

HostBash

Run shell commands

HostRead

Read local files

HostWrite

Write local files

HostEdit

Edit files in place

HostGlob

Find files by pattern

HostGrep

Search file contents

Terminal

$ chucky prompt "Find all TODO comments and create GitHub issues" \
    --allow-possession

# Claude reasons in the cloud...
# Then executes on your machine:

→ HostGlob searching for TODO comments...
→ HostGrep found 12 TODOs across 8 files
→ HostBash gh issue create --title "Implement caching"...
→ HostBash gh issue create --title "Add error handling"...
...

✓ Created 12 GitHub issues from TODO comments

No other platform lets cloud AI control your local machine. This is the future of AI-powered development.

Only on Chucky

The billing problem, solved

Every AI platform makes you build your own usage tracking. We handle it.

Budget Caps

Set $5/day limits per user. Agent sleeps when budget exhausted. No surprise bills, ever.

Agents that sleep until paid

Prepaid Credits

Users top up like an arcade. When credits run out, agent stops. You control the experience.

The arcade token model

Automatic Tracking

Per-user, per-session, per-token. Real-time dashboard shows everything. Export anytime.

Zero metering code

"I can't price my SaaS at $20/month if one power user can burn $50 of compute in a day."

— Every AI developer, before finding Chucky

Token with budget

// Your backend generates tokens with budgets
const token = await createToken({
  userId: 'user_123',
  secret: process.env.CHUCKY_SECRET,
  budget: {
    aiDollars: 5,      // $5 AI spend limit
    computeHours: 2,   // 2 hours compute
    window: 'day'       // Resets daily
  }
});

// When budget exhausted, agent stops automatically
// No overage. No surprise bills. You're in control.

Zero Infrastructure

AI that works while you sleep

Schedule AI tasks. Fire-and-forget jobs. No Redis. No workers. Just cron expressions and a managed queue.

Cron Schedules

Define schedules in chucky.json, deploy with the CLI. Daily audits, weekly reports, monthly reviews.

chucky deploydeploy once

Background Jobs

Fire-and-forget tasks via /incubate API. Get a run ID, move on. Results delivered via callback.

POST /incubateasync execution

Only on Chucky

Managed Queue

Per-user concurrency built-in. Fair scheduling across customers. Cancel pending, monitor runs.

concurrencyKeyauto rate limiting

chucky.json

// Define schedules in chucky.json, deploy with `chucky deploy`
{
  "projectId": "your-project-id",
  "crons": [
    {
      "cron": "0 9 * * 1-5",           // Weekdays at 9am
      "message": "Audit the codebase for security issues",
      "callback": { "url": "https://my-app.com/webhook" }
    }
  ]
}

No polling. No workers. AI automation that just works.

Flexibility

Choose your model, control your costs

Use Claude, OpenAI, or 100+ OpenRouter models. Configure different models for main tasks, subagents, and background jobs.

Anthropic

Claude Opus, Sonnet, Haiku

Best quality, default provider

OpenAI

GPT-5.2, GPT-5, o3, o4

Alternative provider

OpenRouter

100+ models

Cost optimization, free tiers

Configure per-session

const session = client.createSession({
  env: {
    ANTHROPIC_MODEL: 'claude-sonnet-4-5-20250929',
    CLAUDE_CODE_SUBAGENT_MODEL: 'or:mistralai/devstral-2512:free',
    ANTHROPIC_DEFAULT_HAIKU_MODEL: 'or:deepseek/deepseek-chat',
  }
});

See all environment variables

Capabilities

Everything you need to ship AI

Same API handles PR reviews, code explanations, migrations, bug hunts. The agent figures it out.

Only on Chucky

Built-in per-user billing

JWT tokens with embedded budgets. Auto-enforcement at the edge. No Stripe metering code.

Browser + server tools

Define tools that execute in browser OR server. Cloud brain, local hands.

The API can think

Not just execute code — reason through problems, decide what to do, chain multiple steps.

Full codebase context

Your entire repo is loaded. It understands your patterns, your domain, your architecture.

MCP servers & tools

Connect to databases, APIs, external services. Define custom tools in the sandbox.

Multi-turn sessions

Stateful conversations with automatic context management. Claude remembers everything.

Streaming responses

Real-time token streaming for instant feedback. Show partial results as Claude thinks.

Full file & shell access

Read files, write files, run commands, scripts, builds. Not just generate — execute.

Novel Use Cases

What only Chucky enables

These aren't just AI features — they're new categories of apps that require Chucky's unique architecture.

Possession Mode

AI Dev Automation

Run prompts from CLI with possession mode. Claude reasons in the cloud, but HostBash, HostEdit, and HostWrite execute on your machine.

chucky prompt 'Fix all TypeScript errors' --allow-possession

Browser Control

AI Browser Extensions

Build Chrome/Firefox extensions where Claude reads pages, extracts data, fills forms. Tool handlers give AI hands in the DOM.

mcpServers: [{ tools: [...] }]

Built-in Billing

White-Label AI Platform

Deploy once. Bill each client automatically. JWT tokens carry budgets — per-project, per-user, per-session. No Stripe metering code.

createToken({ userId, budget: { aiDollars: 5 } })

Host Tools

Local AI Agents

Build desktop apps (Electron/Tauri) with AI that reads local files, runs shell commands, and edits code — all from natural language.

prompt('Organize my downloads folder')

CLI

CI/CD AI Integration

Add Claude to your pipeline. Review PRs, generate changelogs, run refactoring. Works with GitHub Actions, GitLab CI, any runner.

chucky prompt 'Review changes in this PR'

Cron Jobs

Scheduled AI Automation

Nightly code audits. Weekly report generation. Monthly compliance checks. Define crons in chucky.json, Claude handles the rest.

cron: "0 9 * * 1-5"

Same SDK. Cloud reasoning. Local execution. Entirely new possibilities.

The Real Cost

What you'd have to build yourself

Without Chucky, shipping AI apps means weeks of infrastructure work. We did it for you.

CategoryBuild yourselfWith Chucky

Infrastructure

Container orchestrationSecure sandboxesAutoscalingMonitoring

→

Secure cloud sandboxAuto-scalingZero config

Billing & Metering

Stripe integrationToken trackingRate limitingDashboards

→

JWT budgetsEdge enforcementAuto tracking

Authentication

API key managementMulti-tenancyPermissions

→

HMAC signingProject isolationBuilt-in auth

SDK Development

WebSocket clientStreamingReconnectionMulti-platform

→

Universal SDKStreaming built-inBrowser + Node

AI Integration

API integrationContext managementSessionsTool calling

→

Full Claude CodeMCP serversMulti-turn

Deployment

CI/CD setupEnvironment configSecrets management

→

chucky deployGit workflowInstant rollback

Background Jobs

Redis/Bull setupWorker processesJob monitoringRetry logic

→

Managed queueCron schedulesZero infrastructure

Pricing

Simple, predictable pricing

Choose a plan that fits your needs. All plans include the same features — just more compute time and concurrent sessions.

Hobby

Perfect for side projects

$5/month

~125 hours of compute

10 GB egress included

2 concurrent sessions

Starter

For growing applications

$29/month

~725 hours of compute

50 GB egress included

5 concurrent sessions

Pro

For production workloads

$99/month

~2,475 hours of compute

200 GB egress included

10 concurrent sessions

All plans include

Pay per second of compute
AI costs passed through at cost
Daily, weekly & monthly budgets
Unused credits roll over
Top up anytime for overage
Automatic scaling to zero
Background jobs
Cron schedules

Need more? Top up anytime — compute at $0.04/hour, egress at $0.03/GB

Free tier available

Ship your AI product today

No credit card required. No database. No Stripe metering. No usage tracking code.

Start free Talk to us

Perfect for

Agencies billing clientsIndie devs shipping fastSaaS teams adding AI