LLM inference API for AI agents: OpenAI-compatible chat completions paid per request in USDC on Base. No account or API key. Summarize, classify, extract, translate, generate and chat with gpt-5.4-nano, gpt-5.4-mini, gpt-5.4, gpt-5.5, gpt-4o-mini and more. Misses billed at cost + 4% on actual token usage (upto scheme); deterministic repeats (temperature 0) are cache hits billed at 55% of provider price — loops get cheaper automatically. From $0.001/call, standard OpenAI POST format.