Sakana Fugu

Sakana Fugu presents a multi-agent system through a single model interface. You send a request to one OpenAI-compatible endpoint, and Fugu decides whether to answer directly or assemble a team of specialized models. It manages model selection, delegation, agent communication, verification, and synthesis internally, so your application does not need to implement a separate orchestration layer. The service offers two models. Fugu is positioned for everyday work where performance and latency need to stay balanced, including coding, code review, chatbots, and other interactive workflows. Fugu Ultra coordinates a deeper pool for difficult, multi-step work where answer quality matters more than speed; the product cites uses such as data-science research, paper reproduction, cybersecurity analysis, and literature or patent investigations. Both models are available through the same API, letting teams switch models without changing their integration. Teams can configure which providers or models participate in Fugu’s pool, supporting data, privacy, compliance, or organizational constraints. The API supports Chat Completions and Responses endpoints and works with the standard OpenAI SDK interface. Developers can connect Fugu to custom workflows or install it for the Codex CLI, while client-side timeouts may need adjustment for complex requests.