AI API Request Builder — OpenAI, Anthropic & Gemini curl

This is a deterministic string-construction tool, not an inference engine — it never calls the providers. For each provider it maps your form onto that API's documented request contract, so the command you copy matches the official reference field-for-field. Three things change between providers: the endpoint, the authentication header, and how the system and user messages are shaped.

1. Endpoint and method. OpenAI and OpenAI-compatible providers use POST {base_url}/chat/completions. Anthropic uses POST /v1/messages. Gemini puts the model in the path — /v1beta/models/{model}:generateContent, swapping to :streamGenerateContent when streaming is on.
2. Authentication. OpenAI sends Authorization: Bearer. Anthropic uses x-api-key plus a required anthropic-version: 2023-06-01 header. Gemini uses x-goog-api-key. The key itself is always an environment-variable placeholder, never a literal.
3. Message shaping. OpenAI keeps the system and user turns inside one messages array. Anthropic pulls the system prompt out to a top-level system string (it is not a message role) and requires max_tokens. Gemini nests text under contents[].parts[].text and a separate system_instruction.
4. Parameters.Temperature and token limits map to each API's field name — max_tokens for OpenAI and Anthropic, maxOutputTokens under generationConfig for Gemini. JSON mode becomes response_format (OpenAI) or response_mime_type (Gemini), and is omitted for Anthropic, which has no such flag.
5. Escaping. Your system and user text is JSON-string-escaped, so quotes and newlines never break the body. curl bodies use a single-quoted JSON payload, with any literal single quote escaped for the shell — the copied command runs as-is.

The Python and Node.js tabs each give two variants: a dependency-free raw call (requests /fetch) and the provider's official SDK (openai, anthropic, google-genai). Every snippet carries a comment linking to the exact reference it was built from, and each field maps one-to-one to that reference — so you can trust the copied request without cross-checking three docs sites.

# OpenAI Chat Completions API — https://platform.openai.com/docs/api-reference/chat/create
curl https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "user",
      "content": "Hello!"
    }
  ],
  "temperature": 0.7,
  "max_tokens": 256
}'

# Anthropic Messages API — https://docs.anthropic.com/en/api/messages
curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "claude-sonnet-5",
  "max_tokens": 512,
  "temperature": 0.5,
  "stream": true,
  "system": "You are terse.",
  "messages": [
    {
      "role": "user",
      "content": "Hi"
    }
  ]
}'

Both blocks are produced by the same function the tool uses, and are checked byte-for-byte against hand-reconciled output in the module's verifyWorkedExamples() cross-check. Switch to the Python or Node.js tab in the tool above to see the equivalent SDK code for either example.

• OpenAI — Chat Completions API reference
• Anthropic — Messages API reference
• Google — Gemini generateContent API reference

Endpoint shapes, headers and body fields on this page were last cross-checked against the official references on 2026-07-09. Provider APIs are versioned; the page is reviewed when a referenced endpoint or required header changes.