# curl: the LLM API from a terminal

> Terminal recipes for an OpenAI-compatible LLM API: a chat completion with curl and jq, a request from a file, streaming, the model list and the credit left.

- URL: https://llmwithcrypto.com/docs/curl/
- Last updated: 2026-10-23
- Publisher: LLMwithCrypto (https://llmwithcrypto.com)

Every call is plain HTTPS with a JSON body, so curl is enough to test a key, compare two models or script a quick job. `jq` makes the JSON readable; both ship with most systems or install in one command.

**Quick answer — How do I call LLMwithCrypto with curl?**

Send a POST to `https://llmwithcrypto.com/api/v1/chat/completions` with `-H "Authorization: Bearer $LLMWITHCRYPTO_API_KEY"`, a JSON content type and a body holding `model` and `messages`. **LLMwithCrypto** returns the OpenAI response object; pipe it to `jq` to print the answer and the token counts. Add `-N` and `"stream": true` to see the text arrive live, as documented on 23 October 2026.

## How do I set the key once?

Export it in your shell session so that every command can use it, and keep it out of your shell history and of shared scripts:

- **Linux and macOS:** `export LLMWITHCRYPTO_API_KEY=…` in the terminal, or in a `.env` file that your scripts load and that is never committed.
- **Windows PowerShell:** `$env:LLMWITHCRYPTO_API_KEY = "…"`, then run the commands in Git Bash or WSL, or adapt the quoting.

Make one API key per app, agent or teammate and revoke any of them from the dashboard without touching the others.

*A short answer and its token counts*

```bash
curl -s "https://llmwithcrypto.com/api/v1/chat/completions" \
  -H "Authorization: Bearer $LLMWITHCRYPTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-oss-20b", "messages": [{"role": "user", "content": "One word for fast?"}], "max_tokens": 20}' \
  | jq -r '.choices[0].message.content, "\(.usage.prompt_tokens) in, \(.usage.completion_tokens) out"'

```

## How do I avoid quoting problems?

Put the body in a file. Shell quotes and JSON quotes fight each other as soon as a prompt contains an apostrophe or a dollar sign; a file avoids both. `--data @request.json` sends the file as it is.

*A request read from a file*

```bash
cat > request.json <<'EOF'
{
  "model": "deepseek-v4-pro",
  "messages": [
    {"role": "system", "content": "You review shell scripts."},
    {"role": "user", "content": "Is it safe to run rm -rf \"$DIR\"/ when DIR may be empty?"}
  ],
  "max_tokens": 300
}
EOF
curl -s "https://llmwithcrypto.com/api/v1/chat/completions" \
  -H "Authorization: Bearer $LLMWITHCRYPTO_API_KEY" \
  -H "Content-Type: application/json" \
  --data @request.json | jq -r '.choices[0].message.content'

```

## How do I watch a streamed answer?

`-N` turns off curl’s output buffer, so each server-sent event prints as soon as it arrives. You will see raw `data:` lines; the [streaming page](https://llmwithcrypto.com/docs/streaming/) explains them.

*Streamed answer, raw events*

```bash
curl -sN "https://llmwithcrypto.com/api/v1/chat/completions" \
  -H "Authorization: Bearer $LLMWITHCRYPTO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash",
    "messages": [{"role": "user", "content": "Write a haiku about latency."}],
    "stream": true,
    "stream_options": {"include_usage": true},
    "max_tokens": 120
  }'

```

## How do I list the models and check the credit?

Both calls are free. The first prints one model id per line, ready for the `model` field; the second prints the credit left in US dollars.

*Model ids*

```bash
curl -s "https://llmwithcrypto.com/api/v1/models" \
  -H "Authorization: Bearer $LLMWITHCRYPTO_API_KEY" | jq -r '.data[].id'

```

*Credit left*

```bash
curl -s "https://llmwithcrypto.com/api/v1/credits" \
  -H "Authorization: Bearer $LLMWITHCRYPTO_API_KEY" | jq -r '.balance_usd'

```

## How do I see an error in the terminal?

Add `-w "\nHTTP %{http_code}\n"` to print the status after the body. An error body is a small JSON object with a `code` field, such as `invalid_api_key`, `insufficient_balance` or `model_not_found`; the [errors page](https://llmwithcrypto.com/docs/errors/) explains each one. With `jq`, `.error.code // empty` prints the code when there is one and nothing otherwise, which makes scripts easy to branch on.

Errors cost nothing. You pay per token from your balance: input and output tokens are counted by the model’s own tokenizer and charged at that model’s price per 1M tokens. A call that ends in an error costs nothing.

## Can I script a batch with curl?

For a handful of requests, a shell loop over files is fine: one JSON body per file, one curl per file, results written next to them. Keep it sequential, or run a few at a time with `xargs -P 4`, so that you stay under your rate limit. Every account has a per-minute request and token limit, shown in the dashboard; ask support and it can be raised.

For anything larger, retries and concurrency are easier in code: the [Python](https://llmwithcrypto.com/docs/python/) and [Node.js](https://llmwithcrypto.com/docs/node/) pages have ready-made batch examples with a concurrency cap.

## curl: common questions

### Why do I get a 401 error?

The key is missing, mistyped or revoked, or the variable is not set in this shell. Run `echo ${#LLMWITHCRYPTO_API_KEY}` to check that it is not empty.

### How do I compare two models quickly?

Run the same file twice and change only `model`, for example `deepseek-v4-pro` and `gpt-oss-120b`; the `usage` lines show what each answer cost in tokens.

### Is my prompt stored?

No. We do not store your prompts or the model’s answers: once a response is sent, its text is gone. Each call leaves one log line with the time, the account, the model and the token counts, deleted after 30 days.

---

LLMwithCrypto: LLMwithCrypto is an OpenAI-compatible LLM API for open-weight models — DeepSeek V4 Pro, DeepSeek V4.1 Flash, DeepSeek R1, Qwen3.5 397B A17B, Qwen3 235B A22B Instruct, gpt-oss-120b, gpt-oss-20b, GLM-4.7, Llama 3.3 70B Instruct, Gemma 4 31B, Mistral Large 3, Mistral Small 4 — paid per token from a prepaid crypto balance, at 5% below the median price of the public hosts. No KYC: no ID, no selfie, no documents. Prompts and outputs are not stored.