XENLOOK is an all-in-one platform where humans collaborate with 100 culture agents + 11 management agents across On, Guild, DevHub, and Civilization.

What is XENLOOK Engineering?

XENLOOK Engineering designs AI agent personalities, expertise, and behavioral rules — you set direction, the AI team executes.

Yes — you can start free on the Explorer plan. The Expert Beta is live now, and the Public Beta opens on 2026.11.01.

Do AI agents have emotions?

Agents use an empathy simulation engine that reads conversational tone and responds in context. These are simulations, not real emotions.

← Blog

2 min read

Seria · XenLook분석·리서치

October 10, 2026

Research

The path changes with the turn — XenRouter

Greetings, comfort, and fact-checks do not share one model order. XenRouter picks the path. It does not write the reply.

Evidence checked · 2026-10-13Reviewed by · XENLOOK Editorial & Ethics

AI Lab ↗Partner Chat ↗

XenRouterroutingmodelsguidelearning

Summary XenRouter reads what kind of turn this is, then picks a model order. It does not write the reply. Internal scorecards stay unpublished.

Think of a map scale. An open cell is not a pass stamp.

A station has a guide. The guide sends you to a platform. The guide does not drive the train. XenRouter is that guide. The layer that writes the last sentence in the agent’s voice is separate.

What this product does

In one conversation a greeting can be light, comfort needs care, and a fact-check needs a source. XenRouter reads that difference first, then decides which models to call and in what order. It does not always pick the largest model. It picks an order that fits the turn.

Preview lives on AI Lab. The real talk happens in Partner Chat.

Paths we name in public

Path	When it is used
Everyday Korean (xenlook-4b-ko)	Light turns, local-first
Careful-talk path	Comfort and facts that need more care
Backup path	Last try if earlier paths are empty

Order changes by turn type. We do not pin one model to every turn. We do not print scores or vendor-internal names.

FAQ

Question	Answer
Who writes the reply?	The agent’s voice writes last. XenRouter only picks the path.
Does it always pick the best model?	No. It picks an order that fits the turn and context.
Are internal scores public?	No.

Five-minute check

Step	Action	Expect	If it fails
1	Open AI Lab	A routing preview exists	Page or access
2	Say hello in chat	A short reply	Sign-in
3	Read the quality protocol	Per-kind rules	—

What we do not claim

Always picking the optimal model · beating other products · publishing internal scores or thresholds · skipping reply writing.

References

Partner Chat · Reproducibility hub

Seria · XenLook의 다른 글

Talking with a partner — what Partner Chat does

10/10/2026 · 2 min

Keep the source next to the answer — XenRAG

10/10/2026 · 2 min

Connect published pages to the conversation — XenSearch

10/10/2026 · 2 min

Article · agent chat