Claude Code

Claude Code is Anthropic's terminal coding assistant. It understands your codebase, edits across files and runs commands. It uses the Anthropic Messages format (POST /v1/messages), so you only need to point it at AI Nexus with environment variables.

Official site: https://www.anthropic.com/claude-code

PrerequisitesCopy Anchor Link

• An API key from AI Nexus; see API Keys. If you restricted the key to certain models, include the models you want to use, otherwise you get model_not_found. Use GET /v1/models to check which models the key can use; see Quickstart.
• The base URL has no /v1: Claude Code appends /v1/messages itself. See Anthropic format.
• Pick a model that supports /v1/messages. Pick a model on the model square that supports /v1/messages and use its model ID. Models without that endpoint can't be used in Claude Code; for a terminal tool that uses the Responses API, see Codex CLI.

Install Claude CodeCopy Anchor Link

curl -fsSL https://claude.ai/install.sh | bash

echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.zshrc && source ~/.zshrc

Run claude --version afterwards; if it prints a version number, the install worked.

Set environment variablesCopy Anchor Link

export ANTHROPIC_BASE_URL="https://ai.myaiexp.com"
export ANTHROPIC_AUTH_TOKEN="sk-xxxxxxxx"
claude

ANTHROPIC_AUTH_TOKEN is sent as Authorization: Bearer sk-..., which the AI Nexus /v1/messages endpoint accepts.

Or use settings.jsonCopy Anchor Link

Edit ~/.claude/settings.json (on Windows, %USERPROFILE%\.claude\settings.json) and put the variables under env so you don't have to export them every time:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://ai.myaiexp.com",
    "ANTHROPIC_AUTH_TOKEN": "sk-xxxxxxxx"
  }
}

If the file already has other settings, merge just the env block instead of overwriting the file.

Choose modelsCopy Anchor Link

Claude Code picks models by tier (Opus / Sonnet / Haiku). If the default model names are not in your available list, pick a model on the model square that supports /v1/messages and set these variables to its model ID:

Replace your-model-id and your-fast-model-id in the samples with model IDs from the model square that support /v1/messages (or call GET /v1/models).

export ANTHROPIC_MODEL="your-model-id"  # replace with a model ID from the model square that supports /v1/messages
export ANTHROPIC_DEFAULT_OPUS_MODEL="your-model-id"
export ANTHROPIC_DEFAULT_SONNET_MODEL="your-model-id"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="your-fast-model-id"
export ANTHROPIC_SMALL_FAST_MODEL="your-fast-model-id"

The same works in settings.json:

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://ai.myaiexp.com",
    "ANTHROPIC_AUTH_TOKEN": "sk-xxxxxxxx",
    "ANTHROPIC_MODEL": "your-model-id",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "your-model-id",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "your-model-id",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "your-fast-model-id"
  }
}

You can also type /model inside Claude Code to pick a model. For the full model list, see the model square or GET /v1/models.

VerifyCopy Anchor Link

Start claude, ask anything, then check the console's Usage Logs for the matching entry.

• model_not_found: check that the model name matches GET /v1/models, that the model supports /v1/messages, and whether the key has model restrictions.
• 401: check that the key is correct and not expired or disabled.
• Requests going to /v1/v1/messages: the base URL has an extra /v1; remove it.

See Errors for more troubleshooting.