How to Set Up Claude Code: Install, Configure and Actually Use It

How to Set Up Claude Code: Install, Configure and Actually Use It

By the end of this guide you will have the Claude Code command‑line interface installed on your workstation, authenticated with your Anthropic API key, a small test project initialized with a CLAUDE.md file, and your preferred default settings (model, temperature, output format) saved for everyday use. You will be able to run Claude Code commands from any terminal and see a coherent response from the model. The whole process usually takes between fifteen and thirty minutes, depending on how quickly you can locate your API key and whether you need to create a virtual environment.

More Tutorials & Self-Hosting guides on TechPulseMind →

Prerequisites

If you prefer to keep the installation isolated from your system Python, you can create a virtual environment later in the guide; otherwise be ready to type your password for a sudo‑protected pip install.

Step‑by‑step guide

1. First verify that the basic tools are present and meet the version requirements. Open a terminal and run the version checks for Python and git. This tells you whether you satisfy the software prerequisites before proceeding.

   Explanation: python3 --version prints the interpreter version; git --version shows the git client version.

   python3 --version
   git --version
   

2. Install the Claude Code package using the method recommended by Anthropic. The official documentation typically suggests a plain pip install; you can add the --user flag to avoid needing sudo, or you can omit it for a system‑wide install if you are comfortable entering your password.

   Explanation: This command downloads the latest release from PyPI and places the executable in the appropriate bin directory.

   pip install --user claude-code
   

   If you encounter an “externally‑managed environment” error on a recent Linux distribution, see the troubleshooting section later for a virtual‑environment workaround.

3. Confirm that the installation succeeded and that the executable is on your PATH. The version command should return a string without error.

   Explanation: claude-code --version (or claude --version depending on the exact binary name) queries the installed package for its version number.

   claude-code --version
   

   If the command is not found, locate the install directory with pip show claude-code and add its bin folder to your PATH as described in the verification step.

4. Set up authentication so the CLI can talk to Anthropic’s servers. The safest way is to export your API key as an environment variable in the current shell, or to add it to the configuration file that the CLI reads.

   Explanation: Exporting ANTHROPIC_API_KEY makes the key available to any child process; the CLI will pick it up automatically.

   export ANTHROPIC_API_KEY="sk-ant-…"
   

   Replace the placeholder string with the key you copied from the Anthropic console. To avoid re‑exporting the key in every new terminal, you can add the export line to your shell’s startup file (~/.bashrc, ~/.zshrc, or the equivalent for your shell).

   After setting the variable, run the auth‑status command to verify that the key is accepted.

   claude-code auth status
   

   You should see a message indicating “Authenticated” and possibly the account identifier associated with the key.

5. Create a temporary workspace to test the initialization command. This will generate a CLAUDE.md file that serves as a project‑level manifest for Claude Code.

   Explanation: mkdir makes a new directory, cd changes into it, and claude-code init scaffolds the default files.

   mkdir ~/claude-test
   cd ~/claude-test
   claude-code init
   

   List the directory contents to confirm that CLAUDE.md (and any other expected files) appear.

   ls -la
   

   Open CLAUDE.md with your preferred text editor to inspect the template header that the CLI inserted.

6. Adjust the user‑level settings to match your typical workflow. The CLI reads a JSON (or YAML) configuration file; you can edit it safely with the built‑in edit command.

   Explanation: claude-code config --edit opens the correct configuration file in your default editor, allowing you to change values such as the default model, temperature, and output format.

   claude-code config --edit
   

   Inside the editor, locate the keys default_model, temperature, and output_format. Set them to values that suit your use‑case, for example:

   • "default_model": "claude-3-opus-20240229"
   • "temperature": 0.7
   • "output_format": "markdown"

   Save the file and exit the editor. Then run the show command to verify that your changes were persisted.

   claude-code config show
   

   The output should reflect the exact values you just saved.

7. Run a simple query to ensure end‑to‑end functionality. This step confirms that authentication, configuration, and network connectivity are all working.

   Explanation: claude-code ask sends a prompt to the model and prints the response.

   claude-code ask "Explain the difference between a list and a tuple in Python."
   

   You should receive a coherent answer that mentions mutability, syntax, and typical use cases. No error messages about missing credentials, configuration, or time‑outs should appear.

8. (Optional) Make the Claude Code binary easily reachable and enable shell completion. If you installed with the --user flag, the binary likely resides in ~/.local/bin. Adding that directory to your PATH lets you run the command from any location.

   Explanation: The export line appends the user‑local bin directory to your PATH for the current session; adding it to your shell startup file makes the change permanent.

   export PATH="$HOME/.local/bin:$PATH"
   

   To test completion for Bash, run the completion command and source the output; for Zsh or Fish consult the documentation for the appropriate syntax.

   claude-code completion bash > ~/.bash_completion.d/claude
   source ~/.bash_completion.d/claude
   

   Open a new terminal and type claude-code followed by the Tab key; you should see a list of sub‑commands.

Verification

After completing the steps above, run the following quick checks to confirm that each milestone was reached.

• Tool availability: which claude-code returns a path inside ~/.local/bin (or the system bin directory if you used sudo).
• Version reporting: claude-code --version prints a version string without error.

• Authentication: claude-code auth status shows “Authenticated” and displays your account ID or email.
• Project scaffolding: Inside the test directory, ls lists CLAUDE.md and optionally other default files.
• Configuration persistence: claude-code config show outputs the exact values you set for default_model, temperature, and output_format.
• Live query: claude-code ask "What is the capital of Japan?" returns a short answer (“Tokyo”) with no error output.