Zukko Blog
Guides to Claude Code, Cursor, Codex and the Claude and GPT APIs: setup, token savings and troubleshooting.
LangGraph in Python: a first graph with a model via a gateway
5 min · · агенты
We build a small support graph in LangGraph: two models at different prices, branching, a capped retry loop and a clear picture of how execution works.
Codex pricing on your own key: a per-token cost calculation
6 min · · экономия
Three workload profiles, one formula and a table by model. What drives the Codex bill and how to estimate it in advance.
Memory Bank in Cline: Setting Up Project Memory for the Agent
6 min · · агенты
Memory Bank gives Cline memory between sessions through a set of markdown files. Structure, setup via rules and the token price of that memory.
Codex in Russia: installation and use without a ChatGPT subscription
5 min · · IDE
Codex CLI can be pointed at a gateway through a single config file. Installation, config.toml and a breakdown of the typical access and authorization errors.
Ollama Cloud or an API gateway: what to choose for cloud models
6 min · · сравнение
Ollama Cloud continues the familiar Ollama workflow, but in the cloud. How it differs from a pay-per-token gateway and how to choose for your task.
Vibe coding in 1C: an agent, a configuration dump and checking the result
5 min · · агенты
A working scheme for a 1C developer: files instead of the database, an agent in the terminal, rules in CLAUDE.md and a mandatory check in the Configurator.
Claude API docs in plain terms: the Messages API without the fluff
6 min · · API keys
A short reference to the Claude Messages API: what is required, what is deprecated, how to read the response and errors. Examples target the gateway.
A neural network for solving programming problems: learn, do not copy
4 min · · development
A model can be set up so it does not hand over the answer at once but walks you up a ladder of hints. The prompt, the script and the checking tricks.
How to write your own MCP server in Python and connect it to an agent
5 min · · development
We write an MCP server in Python in about 30 lines, test it without an agent, connect it to Claude Code and watch the agent call the tool.
Cline CLI: the Cline Agent in the Terminal with Your Own Model Provider
5 min · · агенты
Cline works outside the editor too. The console version installs via npm, takes tasks from arguments and pipes, and connects to any compatible provider.
DeepSeek API for free: is there a free tier and what is the minimum cost
5 min · · экономия
We do not promise free access, but the minimum is very low. A formula-based count of how many requests fit into one dollar.
PDF to Excel with a neural network: extract tables without copy-paste
4 min · · автоматизация
A working pipeline: pull a table out of a PDF, ask the model for JSON, write xlsx and reconcile the totals so a silent error does not slip through.
Paying for OpenRouter from Russia: why it is hard and what the options are
5 min · · payments
Why topping up OpenRouter from Russia usually fails and what to do: check payment methods with the service or move to a gateway with SBP and crypto. What to change in code.
Cursor in Russian: interface, AI replies and code comments
6 min · · IDE
Three different questions behind one query: the language of AI replies, of the interface and of code comments. Each covered, with .cursor/rules and a token calculation.
LLM in production: a checklist before launching an AI feature
6 min · · API keys
What to check before releasing a feature built on a language model: keys, limits, timeouts, retries, a fallback model, PII-free logs, cost tracking and regression tests.
AI for a sales team: scripts, emails and call review from transcripts
5 min · · бизнес
Four sales tasks a model can take on: scripts, customer emails, call review and CRM drafts. With a code example and a monthly estimate.
An AI agent for business: tasks, architecture and a realistic budget
6 min · · бизнес
No promises to replace a department. Real agent tasks, a design of model, tools and guardrails, a monthly budget calculation and a pre-launch checklist.
Copilot: Language model unavailable and other errors, what to do
6 min · · errors
What to do when Copilot says Language model unavailable or "We are having troubles", and how to keep working on your own key.
OpenAI API in Ollama: one code path for a local and a cloud model
5 min · · development
Ollama answers requests in the OpenAI format. So one piece of code can target a local model or a gateway just by changing base_url and the model name.
DeepSeek API price: cost of V4 Pro and V4.1 Flash worked through
5 min · · models
What requests to three DeepSeek models cost: the formula, three typical scenarios at 10,000 requests, the role of the cache and a usage-based cost script.
opencode or Qwen Code: which to pick for working with many models
5 min · · сравнение
Both agents are open source and not tied to one vendor. How provider setup, configs, agents and spending differ when you use the same key.
Windsurf AI: what the editor is and how to work with it from Russia
6 min · · IDE
What Windsurf is, how it relates to Cursor and Devin, and what to do when sign-in or payment from Russia fails: editors and agents that take your own key.
How to update Claude Code, roll back a version or remove it completely
6 min · · development
Commands to update, roll back and remove Claude Code by install method, and how to keep your gateway connection settings when reinstalling.
Open WebUI API: sending requests to your models from code
5 min · · API keys
Open WebUI has its own API: the models you see in the chat are reachable from scripts. We show how to get a key, which endpoints to call and where the two keys get mixed up.