Guides to Claude Code, Cursor, Codex and the Claude and GPT APIs: setup, token savings and troubleshooting.
5 min · · агенты
An agent is a model that picks its own actions and repeats a loop until the task is done. The structure, examples and honest limits.
5 min · · сравнение
Two terminal agents with a similar feature set. We compare setup, models and API cost so the choice is not a blind one.
6 min · · сравнение
Both live in the terminal and connect to third-party models, but they are built differently: aider drives git for you, opencode is closer to an agent workbench.
5 min · · development
What LangChain is and which packages it consists of, which tasks it helps with and when the plain OpenAI SDK is enough. We compare code and spending.
6 min · · IDE
Cline reads your project, edits files and runs commands with your approval. Plan and Act modes, connecting your own key, and the cost of a typical task.
6 min · · агенты
Claude Code works not in a chat window but in your project folder: it reads files, edits them and runs commands. What it can do and what it costs in tokens.
6 min · · агенты
Which vibe coding tools you can connect to models through a gateway, how to set them up and what an hour of agent work costs.
5 min · · IDE
Install Claude Code, point it at a gateway with two variables and pay per token by SBP, MasterCard or crypto. Plus the first-run errors and how to fix them.
6 min · · агенты
Without MCP the model invents attributes your configuration does not have. We show what an MCP server for 1C provides, how to install mcp-1c and connect an agent through a gateway.
5 min · · development
An agent can build a one-page site in one session if you brief it well and check the result in a browser. The process and the price, step by step.
5 min · · development
LangGraph does not compete with LangChain, it sits underneath. We cover when a chain or create_agent is enough, when you need a stateful graph, with code on one model.
5 min · · агенты
The model calls nothing itself: it only asks your program to run a function. The whole loop step by step, with Python code and a price estimate.
5 min · · агенты
AGENTS.md is plain Markdown with instructions for the agent: build commands, tests, code style. How Codex discovers these files and what to put in them.
4 min · · development
Claude Code is a terminal agent and needs no GUI. Installing it on a server, authenticating with a Zukko key, working in tmux, and protecting the key.
6 min · · агенты
A subagent handles a side task in its own context and returns only the summary. File format, ways to invoke, and model choice, based on the official docs.
4 min · · development
Tool use lets Claude call functions in your code. The whole loop: tool definitions, parallel calls, errors and limits.
4 min · · агенты
OpenClaw connects to Zukko as a custom provider. We cover the two API options, ready-made openclaw.json configs, and common errors.
4 min · · агенты
Hermes Agent works with Claude and GPT via an API key. Here's which wizard option to pick, which base URL to set, and how to verify the Zukko connection.
7 min · · экономия
We price a month of Claude Code by tokens: three load profiles, a formula with caching, a comparison with subscriptions, and ways to stay on budget.
5 min · · агенты
A minimal AI agent in Python with no frameworks: tools, a call loop, step and budget limits. Full code on the OpenAI SDK and an estimate of what one task costs.
7 min · · агенты
The Model Context Protocol in plain words: how it is built, how to add an MCP server to Claude Code, and why every connected server costs tokens.
6 min · · API keys
An agent stuck in a loop can burn through a balance overnight. A separate key, limits, a model list and proper error handling put a ceiling on spend.
7 min · · агенты
What an agent's bill is made of, a 30-step session worked out on Opus 5.5 and Sonnet 5, where to see actual spend and how to set a ceiling.
3 min · · агенты
One task ate a month's budget — nearly always a loop. The causes, and three ways to protect yourself.