Audience
Companies that need LLM inference under EU data-protection requirements; developers of AI agents and RAG applications
About Heabsy
OpenAI- and Anthropic-compatible inference API from an EU company. The flagship model runs on dedicated GPUs in EIA data centres with zero data retention: prompts and completions are processed in memory only, not stored, not logged, not used for training. Routed open models from third-party providers are available with the same key and clearly labelled. One DPA and one invoice from an EU company. Features: streaming, tool calling, structured output, public DPA and sub-processor list, per-token pricing. Measured on the live system in August 2026: 176 tokens per second per stream, first token in 0.3 seconds.
