Claude Code slash commands worth using

Claude Code runs in the terminal, and part of its controls live in slash commands — lines that start with /. They do not go to the model as a regular prompt. Instead they perform an action inside the client: clearing context, showing state, or running a predefined scenario.

Below is a look at what comes up in daily work, plus a way to add your own commands.

Built-in commands

The set of built-in commands depends on the client version, so the current list is best checked through /help. Here are the ones needed most often.

/clear is more useful than it looks. A long dialog accumulates context that you pay tokens for, and part of it is no longer relevant by the time a new task starts. Resetting before a new job makes answers sharper and cheaper.

/init gathers a project description into a separate file. From then on the model relies on it in each request, and there is no need to re-describe the repository structure every time.

Your own commands

Commands that repeat from project to project are convenient to move into files. Claude Code reads markdown files from a commands directory, and the file name becomes the command name.

Such a file can live inside the project, which makes the command available to the whole team through the repository:

The contents are the instruction the client substitutes for the call:

Review the changes in the current branch.
Point out potential bugs, missing checks,
and places where tests would help.
Answer as a list, without restating the code.

After that, /review-pr runs the prepared scenario. Arguments are passed through a placeholder, which helps when one command serves different inputs.

These files solve a common problem: the same long prompt you have to retype. Moved into a command, it stays consistent and is edited in one place.

Connecting through Zukko

Claude Code can be pointed at a third-party gateway through environment variables. For Zukko these are:

The base URL is set without the /v1 suffix — the client adds the path itself. The variable is named ANTHROPIC_AUTH_TOKEN; with ANTHROPIC_API_KEY Claude Code demands a sign-in to an Anthropic account, so it cannot be used.

The session model is switched with /model, for example /model claude-opus-5-5 — Claude Opus 5.5 at $0.3 input / $0.3 output per 1M tokens. For comparison, the official Claude Opus 5.5 price is $4 input / $20 output per 1M. The current model list is GET https://api.zukko.pro/v1/models.

If /model does not pick up the identifier you need, it is simpler to set the model on the environment side and confirm that the key and address are correct.

Short takeaway

Built-in commands cover session and context control, while your own command files remove the routine of repeated prompts. A good start is /help, /clear, and one or two custom scenarios for tasks that come up every day.

Questions about connecting Claude Code to Zukko go to support: https://t.me/zukkopro?direct↗