Nausicaa is a ledger-driven, Team-first runtime for agents that operate as a topology. Instead of keeping progress only inside a model context, Nausicaa records task requests, progress, results, failures, and recovery boundaries as durable runtime facts. The Main lane can create independent Team members, delegate bounded Worker tasks, exchange authorized A2A messages, and resume work after interruption. Every Lane has its own identity, context, budget, lifecycle, and recovery boundary. The distinctive idea is Teto: a Main-owned auxiliary sensing lane. Main is responsible for acting. Team members provide parallel expertise. Teto watches the execution itself. Teto receives a filtered projection of the run and reasons with its own context, cadence, and budget. It can send a second perspective back to Main at safe boundaries, helping surface: - repeated tool calls and self-reinforcing loops; - stalled progress or budget pressure; - contradictions and missing evidence; - blind spots and alternative approaches; - promising directions the active context may have overlooked. Teto is not a hidden chain-of-thought dump, a post-hoc reviewer, or a fixed Planner/Executor/Reviewer stage. It is an explicit runtime participant with a defined identity, communication path, and lifecycle.

Nausicaa

Nausicaa is a agent harness with an auxiliary sensing lane

Nausicaa is a agent harness with an auxiliary sensing lane