How do I build a request to GPT, Claude, or Gemini API?
Visual builder for LLM API requests: OpenAI, Anthropic, Google. Click fields in the UI, get ready-to-paste code for your app: JSON, cURL, Python SDK, TypeScript fetch.
Every line of code maps exactly to fields in the official API documentation. No magic, it's a UI ↔ JSON spec mapper.
Live preview: estimated tokens, cost at current pricing, context window usage.
Perfect for devs starting with LLM API, prompt prototyping, debugging existing integrations.
How to use it
- Pick a model from the dropdown. Claude Sonnet 4.5 is usually a good start for code-related tasks, GPT-4o-mini for cheap cases, Gemini 2.5 Flash when you need huge context (1M tokens).
- Type the system prompt, the model's instruction (role, tone, output format). For Anthropic this is a separate field, for OpenAI it lands as the first message.
- Add user / assistant messages. Click *"+ Add"*, new block appears. You can simulate full conversation history.
- Set parameters: temperature (0 = deterministic, 2 = wild), max_tokens (response length cap), optionally top_p and stop sequences.
- On the right you see stats (estimated tokens, cost) plus output in 4 formats: JSON, cURL, Python SDK, TypeScript fetch. **Click *"Copy"***: paste into your code.
When this is useful
Six typical situations where this builder gives you a concrete edge:
- First steps with LLM API. You're just starting, don't know the official syntax. Click in the UI, see exactly what the request body should look like.
- Prompt prototyping. Test different system prompts, compare temperature 0.3 vs 0.7, pick the best parameter set. Export code when ready.
- Debug integrations. *"Why does my request return 400?"*. Paste params into the builder, compare generated JSON with what you're sending. Often you'll see missing "max_tokens" field (Anthropic requires it, OpenAI doesn't).
- Migration between providers. Working code for GPT, want to move to Claude. Change model in the builder, see syntax differences (system prompt placement, parameter naming).
- Cost estimation. *"How much will sending 1000 such requests cost?"* Builder calculates live.
- Code snippet for documentation / blog post. Generate cURL example showing the API call for each provider.
Questions and answers
No. Builder runs 100% locally in your browser. It generates code with `$OPENAI_API_KEY` as a placeholder, you never type your real key into our tool. Replace the env var in your own app before running.