Zukko Blog
Guides to Claude Code, Cursor, Codex and the Claude and GPT APIs: setup, token savings and troubleshooting.
MCP server for web search: how to give an agent access to the web
6 min · · агенты
A model only knows what was in its training data. Here is how to connect web search through MCP, how the options differ and how many tokens the results take.
AI for a WordPress site: theme, plugin and edits with an AI agent
5 min · · development
A working process: a local copy of the site, a terminal agent, WordPress rules, code checks and safe storage of the API key.
AI for HTML and CSS: layout from a description or a screenshot
5 min · · development
How to get a model to write markup without junk divs and broken responsiveness: prompt, mockup screenshot input, a Python script and result checks.
Vibe coding a game: building a browser game with an AI agent
4 min · · агенты
Step by step we build a simple JavaScript game with the Claude Code agent: how to phrase the mechanics, how to iterate, which bugs to catch and what a session costs.
A neural-network chat on Gradio: a streaming UI in an hour
5 min · · development
We build a Gradio chat web UI where the reply streams in, you can pick a model and a system prompt, and every answer shows its cost in dollars.
A neural network for Excel macros: VBA code from a task description
5 min · · автоматизация
How to phrase a task for the model to get a working VBA macro, how to test it on a copy of the file and what to do about Run-time errors.
Formatting documents to GOST with a neural network: what AI can and cannot do
5 min · · автоматизация
The model helps with structure, bibliography and captions, but margins, fonts and spacing are better set by a template. A division of labour and a .docx assembly example.
Rate limits in an API: the kinds of limits and a client built for them
5 min · · API keys
What a 429 really means, how RPM differs from TPM and burst, and how to write a client that stays under limits and does not retry what cannot be retried.
Dify: what this platform is and how to connect your own models
6 min · · автоматизация
Dify builds chatbots, workflows and agents in a visual editor. We cover app types, the knowledge base, publishing and connecting models through an OpenAI-compatible API.
Aider: what this AI coding assistant is and how it works with git
6 min · · агенты
Aider is a command-line tool that edits your repository on request and commits every change. Here is how it works and what a session costs.
Creating Excel tables with a neural network: structure, formulas and an XLSX file
4 min · · автоматизация
The model describes the table as JSON, and an openpyxl script builds a finished XLSX with formulas, formats and totals. The schema and the code.
GitHub Copilot in VS Code: setup and connecting your own models
6 min · · IDE
Enable GitHub Copilot in VS Code, then connect your own models through Custom Endpoint with an API key, and see what works without a subscription.
OpenAI Codex: what it is, its parts and how to use it
6 min · · агенты
Codex is a coding agent that reads a project, edits files and runs commands. Its parts, how it differs from ChatGPT and the payment options.
Neural network API key: what it is, where to get it and how not to lose money
5 min · · API keys
A key is a password to your balance, not to a chat. Where keys come from, how to test one, and how limits protect your money.
Vibe coding: how to start and ship a first project in an evening
5 min · · development
A beginner plan: pick a small utility, set up an agent, phrase the task, check the result and keep everything in git.
Vibe coding for free: what you can do without money and what comes next
6 min · · экономия
An honest look at free options for vibe coding, their limits, and the arithmetic of what an hour on cheap models via API costs.
AI for Arduino: how to write sketches with a model and not fry the board
5 min · · development
A model can draft a sketch but cannot see your wiring. How to describe the board, check code before uploading and debug using Serial output.
AI for 1C: how to write code in the built-in language with a model
5 min · · development
A model cannot see your configuration and may invent methods and attributes. Here is how to give them metadata and examples, verify the code and estimate request costs.
Neural network hallucinations: why a model makes things up and how to reduce it
5 min · · development
Made-up facts cannot be removed completely, but they can be made rare and visible. The causes and five measures with working Python code.
Prompt engineering for LLMs: the basics that work with any model
5 min · · development
A good prompt is not a spell but six concrete parts. We show them on one example where both the answer and the request price change.
A neural network for programming online: a browser chat with no install
5 min · · development
You do not need an editor to get started: a browser and a key are enough. Working prompts, model choice and the price of a typical conversation.
Ollama: what it is, what it's for and how it differs from a cloud API
6 min · · models
Ollama runs open language models on your computer and serves them over a local API. What it can and can't do, and when a cloud API is the better fit.
OpenAI API: what it is, which endpoints exist and why the format became a standard
4 min · · API keys
An API lets your code call a model instead of a person typing in a chat. A look at requests, responses and why the OpenAI format is used far beyond the company itself.
Building a website with AI: what it really costs and how to do it
5 min · · development
A step-by-step business-card site built with the Claude Code agent, with a cost estimate, and an honest note on where "free" ends.