Sakana Chat

Sakana Fugu is a multi-agent system that coordinates specialized AI models through a single API. It dynamically orchestrates a diverse pool of models to solve complex, multi-step tasks, reducing integration complexity while aiming for higher performance and efficiency. The system is designed for developers and teams seeking frontier-level orchestration without single-vendor dependency and with adaptable collaboration patterns.

Access to a coordinated pool of specialized models through one API lets Fugu handle model selection and switching for each task, reducing API complexity while improving cost-performance. It provides an optimized way to utilize multiple models and workflows without needing to manually configure each agent or workflow.

Fugu is built for coding, reasoning, and other quality-critical workflows, coordinating expert agents to deliver stronger, more reliable results on complex tasks. It supports flexible agent selection and allows exclusion of specific providers or models to meet data, privacy, and compliance requirements.

The platform emphasizes research-driven coordination for multi-agent intelligence, drawing on recent ICLR 2026 papers TRINITY and Conductor. It presents a model API that coordinates agents rather than relying on hand-designed workflows, and it offers technical reports for deeper exploration.

Use cases

• Autonomous coding and reasoning workflows that require high-quality, multi-step task execution.
• Workflow orchestration across multiple AI agents to improve performance and cost efficiency.