See what your AI is actually doing

You use AI to help write code. Agent Deck shows you what it costs, what it touched, and what it remembers — all in one place.

The AI-native code editor that gets it exactly right. No approximation. No bloat. Just perfect code, every time.

Join the waitlist

See how it works

Trusted development at leading companies

Built by developers at

//

Core features

Your dev brain.
Structured and reusable.

One place for every AI you use.

You probably use more than one AI. One for hard problems. One inside your editor. One for quick questions. Each one keeps its own chat history. None of them share.


So when you hit a limit on one and switch to another, you start over. You explain everything again. It's slow.


Agent Deck keeps every chat from every AI in one place. Switching takes seconds. Nothing gets lost.

See how it works

One place for every AI you use.

You probably use more than one AI. One for hard problems. One inside your editor. One for quick questions. Each one keeps its own chat history. None of them share.


So when you hit a limit on one and switch to another, you start over. You explain everything again. It's slow.


Agent Deck keeps every chat from every AI in one place. Switching takes seconds. Nothing gets lost.

See how it works

A safe box for every AI action.

AI doesn't just write code. It runs commands. It installs things. It changes files. Most of the time that's fine. Sometimes it breaks stuff.


Agent Deck puts every AI action inside a sealed box. If the AI does something wrong, the wrong thing happens to the box. Your computer stays clean.

See the safe box

A safe box for every AI action.

AI doesn't just write code. It runs commands. It installs things. It changes files. Most of the time that's fine. Sometimes it breaks stuff.


Agent Deck puts every AI action inside a sealed box. If the AI does something wrong, the wrong thing happens to the box. Your computer stays clean.

See the safe box

AI that knows your code.

Most AI tools start every chat from zero. You have to explain your project every single time. That costs money. It costs time.


Agent Deck reads your code once and remembers it. Every new chat starts with the AI already up to speed.

How memory works

Set a budget. AI stops there.

Right now, the AI decides how much to spend. You find out at the end of the month. A small task can quietly burn a hundred dollars.


Agent Deck lets you set a limit before the work starts. When the AI hits the limit, it stops. You see what it's spending as it spends — no surprise bill.

See budgets

Set a budget. AI stops there.

Right now, the AI decides how much to spend. You find out at the end of the month. A small task can quietly burn a hundred dollars.


Agent Deck lets you set a limit before the work starts. When the AI hits the limit, it stops. You see what it's spending as it spends — no surprise bill.

See budgets

//

Use cases

One runtime. Every workflow.

  • import { NextApiRequest, NextApiResponse } from 'next'
import { supabase } from '@/lib/supabase'

export default async function handler(
  req: NextApiRequest, 
  res: NextApiResponse
) {
  if (req.method === 'GET') {
    // AI suggestion: Add pagination and filtering
    const { data, error } = await supabase
      .from('products')
      .select('*')
      .order('created_at', { ascending: false })
    
    if (error) return res.status(500).json({ error })
    return res.status(200).json(data)
  }
}

    One place. Every AI.

    Run Claude, Codex, Copilot, Gemini, and OpenCode from one workspace. Every chat lives in a database on your computer. Hit a limit on Claude mid-task? Export the chat, open it in Codex, and pick up where you left off. Nothing gets lost.

    See how chats work

  • export function validateEmail(email: string): boolean {
  const regex = /^[^\s@]+@[^\s@]+\.[^\s@]+$/
  return regex.test(email)
}

export function validatePassword(password: string): boolean {
  // ⚠️ Error: 'lenght' is not a property of string
  if (password.lenght < 8) {  // ← Ligne 8 avec erreur
    return false
  }
  return /[A-Z]/.test(password) && /[0-9]/.test(password)
}

    Safe by default.

    Every AI action runs inside a sealed box. Agent Deck turns your request into a plan, shows you what the AI wants to do, then runs it in a contained space. Let the AI try risky things. Your computer and your files stay untouched.

    See the safe box

  • import Stripe from 'stripe'

const stripe = new Stripe(process.env.STRIPE_SECRET_KEY!, {
  apiVersion: '2023-10-16'
})

// AI suggestion: Add webhook handler
export async function createPaymentIntent(amount: number) {
  try {
    const paymentIntent = await stripe.paymentIntents.create({
      amount: amount * 100, // Convert to cents
      currency: 'usd',
      automatic_payment_methods: { enabled: true }
    })
    
    return { clientSecret: paymentIntent.client_secret }
  } catch (error) {
    // Ghost text: Handle Stripe errors properly
    throw new Error(`Payment failed: ${error.message}`)
  }
}

    AI that knows your project.

    Agent Deck reads your code and builds plain notes about every part – what it does, how it connects. Every AI chat starts with those notes already loaded. Most tools just grep your files and guess. Yours start with a real map. You stop re-explaining the same project every session.

    See how memory works

  • import { validateEmail, hashPassword } from '../utils/auth'

describe('Authentication', () => {
  describe('validateEmail', () => {
    it('should accept valid email addresses', () => {
      expect(validateEmail('user@example.com')).toBe(true)
      expect(validateEmail('test+tag@domain.co.uk')).toBe(true)
    })
    
    it('should reject invalid email addresses', () => {
      expect(validateEmail('invalid')).toBe(false)
      expect(validateEmail('@example.com')).toBe(false)
    })
  })
  
  // AI suggestion: Add test for password hashing
  describe('hashPassword', () => {
    // Ghost text suggestions...
  })
})

    Spend with intent.

    Agent Deck sets a limit on the AI before it spends. Set the cap per AI, per project, or per chat. The AI stops when it hits the limit. You see the spending live on every chat and on a dashboard. No more bills that catch you off guard.

    See budgets

Multi-provider chats

Sandboxed agents

RAG indexing

Budget gating

One place. Every AI.

Run Claude, Codex, Copilot, Gemini, and OpenCode from one workspace. Every chat lives in a database on your computer. Hit a limit on Claude mid-task? Export the chat, open it in Codex, and pick up where you left off. Nothing gets lost.

See how chats work

//

Benefits

Why people pick Agent Deck.

Real cost. No middleman.

You pay the AI what it actually costs. No markup. No subsidy. The bill is between you and the AI.

No rate limits.

When one AI cuts you off, switch to another and keep going. No waiting. No begging for higher caps.

Nothing gets forgotten.

Your chats, your decisions, the project context — all of it survives restarts, AI changes, and tool swaps.

Your chats stay yours.

Every chat lives on your computer first. Cloud sync only happens if you turn it on, and only for what you pick.

Works with what you already use.

Claude, Codex, Copilot, Gemini, OpenCode. Agent Deck runs all of them. You don't have to pick one.

Experiment without breaking things.

Every action runs inside a safe box. Try the risky thing. If it breaks, your computer doesn't.

//

Pricing

Pay once for the runtime.
Add cloud as you need it.

Waitlist

$25

one-time

Perfect for trying Exact. All core features included, forever.

Join waitlist

Features

Unlimited local projects

AI autocomplete (basic)

Code chat assistant

50+ languages support

Terminal integration

Runtime

Popular

$30

$30

one-time

For developers who need advanced features and unlimited AI requests.

Coming soon

Features

Multi-provider chats (Claude, Codex, Copilot, Gemini, OpenCode)

Automated sandboxing

RAG indexing and auto-docs

Token budgeting and cost gating

Local SQLite chat store

Export and import across providers

Agent Pro

$15-25/mo

For teams that need dedicated support, custom deployment, and advanced security.

Reserve pricing

Features

Everything in Runtime

Chat search (try 30 days free)

ChatSync across devices and team (mobile coming soon)

Cost Sync (shared team budgets and visibility)

SSO and SAML

Self-hosted deployment

Custom AI model training

Waitlist

$25

one-time

Perfect for trying Exact. All core features included, forever.

Join waitlist

Features

Unlimited local projects

AI autocomplete (basic)

Code chat assistant

50+ languages support

Terminal integration

Runtime

Popular

$30

$30

one-time

For developers who need advanced features and unlimited AI requests.

Coming soon

Features

Multi-provider chats (Claude, Codex, Copilot, Gemini, OpenCode)

Automated sandboxing

RAG indexing and auto-docs

Token budgeting and cost gating

Local SQLite chat store

Export and import across providers

Agent Pro

$15-25/mo

For teams that need dedicated support, custom deployment, and advanced security.

Reserve pricing

Features

Everything in Runtime

Chat search (try 30 days free)

ChatSync across devices and team (mobile coming soon)

Cost Sync (shared team budgets and visibility)

SSO and SAML

Self-hosted deployment

Custom AI model training

Waitlist

$25

one-time

Perfect for trying Exact. All core features included, forever.

Join waitlist

Features

Unlimited local projects

AI autocomplete (basic)

Code chat assistant

50+ languages support

Terminal integration

Runtime

Popular

$30

$30

one-time

For developers who need advanced features and unlimited AI requests.

Coming soon

Features

Multi-provider chats (Claude, Codex, Copilot, Gemini, OpenCode)

Automated sandboxing

RAG indexing and auto-docs

Token budgeting and cost gating

Local SQLite chat store

Export and import across providers

Agent Pro

$15-25/mo

For teams that need dedicated support, custom deployment, and advanced security.

Reserve pricing

Features

Everything in Runtime

Chat search (try 30 days free)

ChatSync across devices and team (mobile coming soon)

Cost Sync (shared team budgets and visibility)

SSO and SAML

Self-hosted deployment

Custom AI model training

//

FAQ

Questions? We've got answers.

Does Agent Deck replace Claude or Copilot?

Which AI tools work with it?

What runs on my computer and what runs in the cloud?

How does the safe box work?

Does it work without internet?

What if I want to share with others?

Own your AI context. Starting now.

Join the paid waitlist for $25. Lock in lifetime discounts on every cloud tier, plus first access to the runtime when it ships.

Join waitlist – $25