Kimi Code

Kimi Code CLI is Moonshot AI’s terminal AI agent. In its own words, it runs in the terminal and helps with software development and everyday terminal work: reading and modifying code, running shell commands, searching files, fetching web pages, and planning its next steps based on feedback. It is written in TypeScript and distributed via npm.

Don’t install the old version

The old Python Kimi CLI (repository MoonshotAI/kimi-cli) is archived and no longer maintained; existing installations will stop working. It has been replaced by Kimi Code CLI. The PyPI package named kimi-code is a legacy alias of the old CLI, not the new Kimi Code CLI. Install it as shown below.

Prerequisites

• A NoviaHub account with an API key. See API keys.
• Funds in your account. See Wallet and top-ups.

Install Kimi Code CLI

macOS / Linux

The install script is the recommended method and doesn’t need Node.js installed first:

curl -fsSL https://code.kimi.com/kimi-code/install.sh | bash

Windows

Before the first launch, install Git for Windows; Kimi Code CLI uses its bundled Git Bash as the shell. Then run in PowerShell:

irm https://code.kimi.com/kimi-code/install.ps1 | iex

npm / Homebrew

Installing with npm needs Node.js 22.19.0 or later:

npm install -g @moonshot-ai/kimi-code
# or
brew install kimi-code

Configure NoviaHub

In Kimi Code CLI, a “provider” decides which protocol to use and where to connect; a “model” is an alias defined on top of a provider. The steps below add NoviaHub as a provider that uses the OpenAI Chat Completions protocol.

1. Store the key in the environment variable NOVIAHUB_API_KEY.

   macOS / Linux

   # macOS uses zsh by default; with bash on Linux, use ~/.bashrc instead of ~/.zshrc
   echo 'export NOVIAHUB_API_KEY="sk-..."' >> ~/.zshrc
   source ~/.zshrc

   Windows

   setx NOVIAHUB_API_KEY "sk-..."

   setx does not affect the current window; close the terminal and open a new one.

2. Edit the config file ~/.kimi-code/config.toml. It is created automatically the first time you run kimi; if it isn’t there, create it. Add:

   ~/.kimi-code/config.toml

   default_model = "noviahub/deepseek-v4-flash"
   
   [providers.noviahub]
   type = "openai"
   base_url = "https://noviahub.com/v1"
   api_key_env = "NOVIAHUB_API_KEY"
   
   [models."noviahub/deepseek-v4-flash"]
   provider = "noviahub"
   model = "deepseek-v4-flash"
   max_context_size = 1000000

   Setting	Purpose
   default_model	The model alias new sessions use. It must be defined in a [models."..."] table below. If the file already has a default_model, change its value instead of adding a second line.
   type	The protocol. openai means OpenAI Chat Completions.
   base_url	The API address, including /v1: https://noviahub.com/v1.
   api_key_env	Which environment variable to read the key from. It is read again on every request.
   [models."noviahub/deepseek-v4-flash"]	The model alias. The name contains /, so it must be quoted.
   provider	Which provider to use; matches [providers.noviahub] above.
   model	The model ID sent to NoviaHub. It must match the model ID on NoviaHub exactly.
   max_context_size	Required. The model’s context length in tokens, shown as Context on the model’s page under Models & Pricing (deepseek-v4-flash listed 1,000,000 on 2026-09-28).

3. Check the config file:

   kimi doctor

   It only validates config.toml and tui.toml; it doesn’t start the interface or change either file.

4. Run kimi. New sessions use the model set in default_model.

Shell environment variables are not read automatically

Kimi Code CLI reads credentials only from its config file. Running export KIMI_API_KEY or export OPENAI_API_KEY in a terminal gives no provider a key. This page names the variable explicitly with api_key_env in the config, so the key stays out of the file. If that variable is unset or empty, the request fails with an error that names the variable.

Note

The code samples under API access on NoviaHub’s model pages name the variable NEW_API_KEY. Only the name differs; pick one and use it consistently. If you use NEW_API_KEY, change api_key_env to match.

Verify

1. Send one message in non-interactive mode; the answer is printed straight to the terminal:

   kimi -p "Introduce yourself in one sentence"

2. You can also run kimi provider list to see the configured providers.

3. Check the call and its cost under Usage Logs in the NoviaHub console.

Switch and add models

• Add a model: add another [models."<alias>"] table with provider set to noviahub, model set to the NoviaHub model ID and max_context_size set to that model’s context length.
• Switch models: run /model in the interface, or pass an alias at launch, for example kimi -m noviahub/deepseek-v4-flash.
• Which models: type = "openai" uses the Chat Completions API, so pick models that carry the Chat label under Models & Pricing. Every model on the site carried it on 2026-09-28.
• Use the Responses API: Kimi Code CLI also supports type = "openai_responses" (the OpenAI Responses API, configured the same way as openai). Only models whose page shows the Response label work with such a provider.

Notes

• /login is only for signing in to the Kimi Code managed service or entering a Kimi Platform API key. It is not how you connect NoviaHub; edit config.toml as above.
• Set exactly one of api_key and api_key_env. api_key writes the key into the config file in plain text, which isn’t recommended.
• To keep the data directory elsewhere, set the environment variable KIMI_CODE_HOME; the config file is then $KIMI_CODE_HOME/config.toml.

References

Checked on 2026-09-28:

• Getting started (install, first launch): https://moonshotai.github.io/kimi-code/en/guides/getting-started
• Providers and models: https://moonshotai.github.io/kimi-code/en/configuration/providers
• Config file fields: https://moonshotai.github.io/kimi-code/en/configuration/config-files
• kimi command reference (-p, -m, doctor, provider list): https://moonshotai.github.io/kimi-code/en/reference/kimi-command
• Kimi Code repository: https://github.com/MoonshotAI/kimi-code
• Old Kimi CLI (archived): https://github.com/MoonshotAI/kimi-cli