A powerful coding add-on for VS Code.

Keep the quality. Cut the cost.

Bring your own key and build with bigger models, without the token anxiety.

Install CodeCook

Why I built CodeCook

I wanted to spend more time building. Not worrying about tokens.

Long chats resend everything you already paid for: every old message and every tool output, on every turn. The bill keeps growing even when the work doesn’t.

CodeCook summarizes the older turns, keeps your plan in reach, and shows what each turn will cost.

After CodeCook’s smart compression, 800,787 tokens became 14,857.

Features

Less repetition. A plan you keep. A cost you can see. A diff you review.

Less repetition

Before

800,787 tokens. Earlier turns, tool output, and repeated file context.

After

14,857 tokens. Summary kept. Recent turns still sent.

Keep the plan

plan.md

Auth stays on the existing provider.

Responses follow the current schema.

Keep the existing database tables.

See the cost

Every turn shows what was sent, what was saved, and where your month stands.

CodeCook details with the monthly cap, summary count, Summarize chat, and Savings report. — Monthly cap, summaries, and the savings report.

Review the diff

Every change shows as a diff before you accept it.

− sendFullHistory();

+ sendCompactHistory();

Control

Your keys. Your costs. Your code.

Your key

Claude, GPT, Google Cloud, AWS, or Azure. Stored on this machine.

Your estimate

Token fields and spend caps are estimates, not the provider bill.

Your diff

Open it, keep the turn, or revert it.

Who it’s for

Built for developers who want more from their AI budget.

Freelancers

Spend less time re-explaining a client project and more time delivering the work.

Indie developers

Keep building without managing a long conversation by hand every time the context grows.

Startups

Give the team a shared way to see estimated usage and review the diff before it stays.

FAQ

Straight answers.

Does CodeCook include AI model usage?

No. You bring a key for a supported provider. Model usage and charges depend on that account, the model, and the configuration.

Can I use bigger models?

You can select a supported model on the provider you connect. Availability, limits, and price stay with that provider. Summarizing older chat is meant to leave more of the prompt for the current work. It does not change what the provider allows.

How does CodeCook help reduce token usage?

Summarize chat compresses older history, and later turns can send that summary instead of the full earlier text. The token panel compares history-if-full with history sent. The reduction depends on the conversation, the model, and when you summarize.

Does reducing tokens affect code quality?

A summary can leave details out. CodeCook keeps the recent turns and a plan you can reopen. Read the diff before you keep the change.

Do I still pay the AI provider?

Yes. With your own key, you pay that provider. Check their current prices and usage records. Figures inside CodeCook are estimates.

Does CodeCook work with my current coding tools?

It runs in VS Code 1.85 or newer, Cursor, and Antigravity. The key form supports Google, ChatGPT, Claude, Azure OpenAI, and AWS. It is not a plug-in for every AI coding product.

Is my API key stored on your servers?

Provider keys are stored in VS Code Secret Storage on this machine. CodeCook does not run a Makayis backend for those keys. A turn sends the key to the provider you selected. Google sign-in sends the desktop client secret to Google when exchanging the token. This marketing site uses Firebase Analytics, which is separate from the extension. Details are on the privacy page.

How do I get started?

Open the extension folder, install dependencies, compile, and press F5. Sign in, then paste a key for a supported provider. The steps are in Install CodeCook.

Your next coding session can be better.

Summarize the old chat. Keep the plan. Read the diff.