Write code using Claude Code

Olares version: 1.12.7+▾

Olares 1.12.7+Olares 1.12.6

Claude Code is an AI coding assistant that helps you write, test, and manage code using natural language. On Olares, this command-line interface runs inside a browser-based terminal equipped with a pre-configured Ubuntu development environment.

Learning objectives

In this guide, you will learn how to:

• Install the Claude Code app from the Olares Market.
• Connect the Claude Code CLI to a model using an Anthropic subscription or a local model.
• Execute basic and advanced natural language coding workflows.
• Manage software dependencies securely.

Prerequisites

Before you begin, you need:

• An Olares device running Olares 1.12.7 or later, with sufficient disk space and memory.

• Access to Router for local model connections.

• Additional requirements depend on how you connect:

  Connection method	Requirements
  Anthropic subscription	An active Claude Pro or Max subscription
  Local model	Qwen3.8-27B (llama.cpp) installed from Market

Optional: Use a different model

Install another chat model from Market, or create a model instance with Engine Base apps. Then select it as the default chat model on Router's Default models page. Clients configured with default-chat use the new default without changing their model settings.

To keep a client on one specific model, use the full model name from Router's How to call this model window instead.

Model compatibility

Some local models, including Qwen3.5-9B-GGUF Q4_K_M (llama.cpp), use chat templates that are incompatible with Claude Code's system-message ordering. If you encounter a system-message error, switch to another model. See the FAQ for details.

Install Claude Code

1. Open Market, and search for "Claude Code".

   

2. Click Get, and then click Install. Wait for the installation to finish.

Connect to a model

Start the Claude Code CLI and connect it to a language model. Choose one of the following connection methods.

Connect using an Anthropic subscription

Use this method if you hold an active Claude Pro or Max subscription.

1. Open the Claude Code CLI from the Launchpad.

2. Enter the following command:

   claude

3. Select a terminal theme such as Dark mode, and then press Enter.

4. Select Claude account with subscription as the login method, and then press Enter. A browser window opens for sign-in. If the browser fails to open, click the provided URL to sign in manually.

   

5. Complete the sign-in flow in your browser, and then copy the authentication code.

6. Return to the terminal, paste the code, and then press Enter to complete the login.

7. Review the Accessing workspace: /opt/data security prompt, and then select Yes, I trust this folder.

   The Terminal User Interface (TUI) opens automatically.

   

Connect using a local model

Use this method to run Claude Code locally. This example uses the model app Qwen3.8-27B (llama.cpp).

How model connections work

On Olares 1.12.7 and later, AI clients connect through Olares Router. Router provides the Base URL and routes default-chat to the model selected on its Default models page.

This guide uses Qwen3.8-27B (llama.cpp) as the default chat model. For connection formats and API key requirements, see Connect AI apps.

1. Open Router from Launchpad. On Default models, set Qwen3.8-27B (llama.cpp) as the default chat model.

2. Go to LLM, find Qwen3.8-27B (llama.cpp), and click View connection example on its row.

3. In How to call this model, select Apps in Olares and copy the Base URL, then remove the trailing /v1 for the Anthropic-compatible client. For example, use https://router.<your-olares-domain>. The client appends /v1/messages.

4. For the model name, use default-chat.

   Selecting a model in your app

   default-chat is a routing name and does not appear in model lists loaded from Router. If the app you are connecting to Router loads a model list, add default-chat manually. If the app only supports selecting a listed model, select the full model name shown in Router's How to call this model window.

5. For the API key, leave the field empty. If the app you are connecting to Router requires a value, enter olares.

6. Open Olares Settings, and then go to Applications > Claude Code > Manage environment variables.

7. Specify the following environment variables:

   • ANTHROPIC_AUTH_TOKEN: Enter olares.
   • ANTHROPIC_BASE_URL: Enter the Base URL you copied from Router. For example, https://router.<your-olares-domain>.
   • ANTHROPIC_MODEL: Enter default-chat.

8. Click Apply. Wait about 10 seconds for the container to restart.

9. Open the Claude Code CLI from the Launchpad, and then enter claude in the terminal to start your session.

Use Claude Code

All project work happens in the /opt/data directory, which serves as $HOME in the container. This directory persists your files across app restarts.

The following examples demonstrate how to interact with Claude Code to complete everyday development tasks.

Run basic queries

1. In the Claude Code CLI, enter the following command:

   claude

2. Choose the text style that looks best with your terminal, such as dark mode.

3. Review the security notes, and then press Enter to continue.

4. Review the Accessing workspace: /opt/data security prompt, and then select Yes, I trust this folder to grant Claude Code read, edit, and execute permissions.

5. (Optional) In the TUI, run the /clear command to start a new session with empty context.

   

   Switching between modes

   If you switch between remote and local models, run /clear in Claude Code first before starting a new session. This prevents context from the previous model from affecting the new workspace.

6. Describe your task in natural language. For example:

   List the files in the current directory

   The assistant automatically executes the necessary internal commands to explore the directory and returns a detailed list of your files.

   

7. Review the results.

Build a full-stack project

Claude Code creates multi-service projects, runs tests, and verifies end-to-end integrations.

The following example demonstrates how to build a lightweight "Hello Olares" web application using a single Node.js Express server to handle both the backend API and the frontend display.

1. In the Claude Code TUI, enter the following prompt:

   Create a simple full-stack "Hello Olares" application in a new directory called `hello-olares`.
   
   Please do the following:
   1. Initialize a Node.js project and install the `express` package.
   2. Create a backend API (`server.js`) that runs on port 3000 and has a single endpoint `/api/message` returning `{"message": "Hello Olares!"}`.
   3. Create a frontend (`public/index.html`) with vanilla JavaScript that fetches the message from the API and displays it on the screen. Configure the server to serve this static directory.
   4. Start the server in the background, use `curl` to verify the `/api/message` endpoint works, and then stop the server cleanly.

2. Wait for Claude Code to process the prompt. The assistant automatically initializes the project, installs Express, writes the code, starts the server, and performs a live curl integration check.

3. When the assistant prompts you for permission to proceed, select Yes, and don't ask again.... You might need to approve several prompts for different types of actions.

4. Review the final summary report returned by the assistant. It outlines the newly created project structure, the configured backend API, the frontend setup, and the successful curl test results.

   

Manage security and development environments

The Claude Code container operates under strict least-privilege settings to ensure security.

The main process and all executed commands use a non-root user (UID/GID 1000). The container disables allowPrivilegeEscalation and drops all Linux capabilities. Consequently, administrative commands like sudo and apt install are unavailable.

Review pre-installed development tools

Before you install additional software, review the tools already included in your workspace. The container image is based on Ubuntu 24.04 and comes with many common development tools pre-installed.

The following table lists the key categories and examples.

Category	Included tools
Languages and runtimes	Python 3, Node.js, Go, Rust, Java (OpenJDK 21), Ruby, PHP 8.3, Lua, Perl, SQLite
Build tools	build-essential, cmake, ninja-build, clang, pkg-config, common -dev headers
CLI utilities	git, git-lfs, curl, wget, jq, yq, openssh-client, unzip, zip, rsync, tmux, htop, shellcheck
Database clients	postgresql-client, mysql-client, redis-tools

INFO

The ripgrep (rg) utility is intentionally excluded to prevent conflicts with Claude Code's native search behavior.

Install additional software

If your project requires tools or libraries beyond the pre-installed ones, you must manage them within the container's security boundaries.

What you cannot install yourself

If your project requires a system‑level library (e.g., libpq-dev, ffmpeg, libssl-dev), you cannot install it directly. These dependencies must be added to the base container image by the application maintainer.

What you can install in your workspace

Inside your writable directories (primarily /opt/data), you can install project‑level dependencies without root privileges using common tools:

• Python: Create a virtual environment and use pip. For example:

  python3 -m venv .venv
  source .venv/bin/activate
  pip install <package>

• Node.js: Use npm inside your project folder. For example:

  npm install <package>

• Rust/Go (or other compiled languages): Install binaries to a user‑writable path. For example:

  cargo install --root ~/.local <package>   # Rust
  go install <package>@latest               # Go (installs to ~/go/bin)

INFO

The container preconfigures the environment variable PIP_BREAK_SYSTEM_PACKAGES=1. While this permits system‑wide Python package installations, using a virtual environment is recommended to keep your workspace clean and reliable.

FAQs

System message must be at the beginning.

This error occurs when the model's chat template only accepts system messages at the beginning of a conversation. Claude Code can send additional system messages later in the conversation. A template with this restriction rejects those messages, causing HTTP 500 errors and repeated retries.

claude: command not found

Wait a few moments for the init container to finish installing Claude Code. Verify that the $HOME/.local/bin directory exists in your system PATH.

OAuth or install script fails

Verify your Olares cluster's outbound network connection. The init container requires internet access to download dependencies from https://claude.ai/install.sh.

Missing language or library

Determine if the missing tool is a system-level dependency:

• System-level dependencies: You cannot install these yourself. The app maintainers must add them to the base image. If you need a system-level library that is not currently available, submit a GitHub Issue to request it.
• User-level dependencies: Use venv, npm install, or similar local tools to install them.

Learn more

• Set up OpenCode as your AI coding agent
• Claude Code official documentation