# Claude Code (/docs/code-cli/claude-code)

[Claude Code](https://code.claude.com/docs) works with OhMyGPT out of the box. Set two environment variables and it runs on your OhMyGPT key, with every Claude model and the rest of the catalog one `/model` command away. Billing goes to your OhMyGPT balance, and you can test the whole setup for free with `ohmygpt/mock`.

## Configure [#configure]

Set the variables, then start Claude Code from the same shell:

```bash
export ANTHROPIC_BASE_URL="https://api.ohmygpt.com"
export ANTHROPIC_AUTH_TOKEN="$OHMYGPT_API_KEY"
export ANTHROPIC_MODEL="anthropic/claude-sonnet-5-5"
claude
```

- `ANTHROPIC_BASE_URL` has no `/v1`. Claude Code appends `/v1/messages` itself.
- `ANTHROPIC_AUTH_TOKEN` sends your key as `Authorization: Bearer`. `ANTHROPIC_API_KEY` (sent as `x-api-key`) works too, but Claude Code asks you to approve it once.
- `ANTHROPIC_MODEL` is any model ID from [Choose a model](#choose-a-model).

To keep the setup across terminals, put the same values in the `env` block of `~/.claude/settings.json`:

```json title="~/.claude/settings.json"
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.ohmygpt.com",
    "ANTHROPIC_AUTH_TOKEN": "sk-...",
    "ANTHROPIC_MODEL": "anthropic/claude-sonnet-5-5"
  }
}
```

<Callout type="warn">
  Keep the key out of a project's `.claude/settings.json`, because that file is committed with the repository. Use `~/.claude/settings.json` or a git-ignored `.claude/settings.local.json`.
</Callout>

## Check the connection [#check-the-connection]

Send the request Claude Code sends, to the free test model `ohmygpt/mock`:

```bash
curl https://api.ohmygpt.com/v1/messages \
  -H "Authorization: Bearer $OHMYGPT_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model": "ohmygpt/mock", "max_tokens": 64, "messages": [{"role": "user", "content": "ping"}]}'
```

```json
{
  "id": "chatcmpl-ohmygpt-mock",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "This is a mock reply from OhMyGPT. Your integration works; switch to a real model when ready."
    }
  ],
  "model": "mock",
  "stop_reason": "end_turn",
  "usage": { "input_tokens": 19, "output_tokens": 17 }
}
```

Then start `claude` and run `/status`. The **Anthropic base URL** line shows `https://api.ohmygpt.com`, and the credential line names `ANTHROPIC_AUTH_TOKEN`.

You can also run Claude Code itself on the mock with `claude --model ohmygpt/mock`. When Claude Code sends tools, the mock calls the first one and then replies with text, so you see one full turn of the agent loop at no cost. The mock is available once your account has been topped up.

## Choose a model [#choose-a-model]

List the models you can use with Claude Code:

```bash
curl -s https://api.ohmygpt.com/v1/models \
  | jq -r '.data[] | select(.available and any(.protocols[]?; .id == "anthropic-messages")) | .id'
```

Prices for each are on the [Models](https://www.ohmygpt.com/models) page.

Claude Code is built for Claude, and `anthropic/claude-*` models give you the full experience. Other models in the list, such as `zhipu/glm-5.3`, run Claude Code too, including its tool calls and thinking settings.

Switch models inside a session with `/model <model id>`. Claude Code's built-in choices (`opus`, `sonnet`, `haiku`, `fable`) also work, because OhMyGPT accepts the Claude model names they send, such as `claude-sonnet-5-5`.

Claude Code runs some background work, such as session titles, on its Haiku-class model. To choose that model too, set:

```bash
export ANTHROPIC_DEFAULT_HAIKU_MODEL="anthropic/claude-haiku-5-5"
```

To see OhMyGPT's Claude models in the `/model` picker, set `CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1`. Claude Code then reads `GET /v1/models` at startup and lists every ID that contains `claude` or `anthropic`.

## GitHub Actions [#github-actions]

[claude-code-action](https://github.com/anthropics/claude-code-action) sends its `anthropic_api_key` input as `x-api-key`, which OhMyGPT accepts. Store your OhMyGPT key as a repository secret and set the base URL in `env`:

```yaml
env:
  ANTHROPIC_BASE_URL: https://api.ohmygpt.com

steps:
  - uses: anthropics/claude-code-action@v1
    with:
      anthropic_api_key: ${{ secrets.OHMYGPT_API_KEY }}
```

## Troubleshooting [#troubleshooting]

| Symptom | Cause | Fix |
|---|---|---|
| Every request returns `404 page not found` | `ANTHROPIC_BASE_URL` ends in `/v1`, so requests go to `/v1/v1/messages` | Set it to `https://api.ohmygpt.com` |
| `401` with `Invalid API key` | The key is mistyped or was deleted | Copy a key from [API keys](https://www.ohmygpt.com/keys) |
| Login screen at startup | Claude Code found no credential | Export `ANTHROPIC_AUTH_TOKEN` in the shell that runs `claude`, or set it in `~/.claude/settings.json` |
| `ANTHROPIC_API_KEY` is set but not used | The key was declined at Claude Code's approval prompt | Use `ANTHROPIC_AUTH_TOKEN`, or enable the key under `/config` → **Use custom API key** |
| `404` saying the model does not exist | The ID is misspelled | Copy the ID from `GET /v1/models` |
| `402` | The balance or the key's spending limit is used up | Top up at [Billing](https://www.ohmygpt.com/billing) |

Every response carries an `X-Request-ID` header, and [Usage](https://www.ohmygpt.com/usage) shows what each request cost.