LLM API Request Builder
Build OpenAI Responses, Anthropic Messages and Gemini Interactions examples locally, without sending requests or collecting keys.
Generate examples only. No request is sent and no API key is collected. Each field accepts up to 32 KiB UTF-8.
Build a request
Generated code
Examples require a provider SDK for JavaScript/Python and an environment variable in your own environment. Running copied code sends its content to that provider and may incur charges. Streaming examples print raw events; function-call handling is yours to implement.
Responses API documentation · Verified 2026-10-13
Choose a provider and enter user input.
Three APIs, explicit differences.
OpenAI Responses uses input and instructions; Anthropic Messages separates system from messages and requires max_tokens. Gemini Interactions uses input, system_instruction and generation_config. These examples use the reviewed provider format, not a legacy chat completion or generateContent wrapper.
Copy a request; keep control.
DevXTools never executes the example. Set the indicated environment variable outside this site and install the official SDK if you choose Python or Node.js. cURL output targets a POSIX shell and safely quotes prompt data. No API-key field, storage, proxy or provider connection exists here.
Bounded, reviewed support.
Use up to 32 KiB UTF-8 per field and eight function definitions. Response schemas support a conservative, bounded subset; unsupported or unverified combinations fail before code generation. Instructions and schema descriptions remain inert. Provider availability and SDK compatibility still need checking in your own environment.
Final response versus tool arguments.
A response schema shapes the model’s final answer. A tool schema describes arguments for a function your application may later handle. Generating a definition does not implement or execute that function. OpenAI and Gemini examples opt out of response storage; no claim about provider-wide retention is implied.