Execution Modes

Every agent run uses one of three execution modes: Auto, Fast, or Deep.

Auto mode

Auto (default for Internal agents) lets Auteryn classify each incoming message and pick Fast or Deep automatically. Use Auto when you want the platform to balance speed and depth without manual tuning.

Fast mode

Fast is optimized for speed. It’s ideal for:

Short Q&A conversations
Simple lookups and retrievals
Quick summaries
Customer-facing chat where response time matters
High-volume interactions

Responses are typically returned in a few seconds. Fast mode uses fewer credits per run.

Deep mode

Deep is optimized for depth and accuracy. It’s ideal for:

Complex, multi-step tasks (“write a full test suite for this module”)
Long-form research and synthesis
Tasks that require planning and multiple integration calls
Code generation, debugging, architecture review
Anything where quality matters more than speed

The agent spends more time reasoning before and during the task. Deep mode uses more credits and gives better results on complex work.

How mode is selected

Scenario	Mode used
Agent type is Customer-facing	Always Fast
Agent type is Internal + Auto enabled	Auteryn picks Fast or Deep per message
Agent type is Internal + Fast or Deep set	That mode for every message

For Internal agents, set the agent’s mode in its Model section, under How hard it thinks. Loops, schedules, webhooks and API runs use that mode, and a new chat starts from it.

In a chat, the model control next to the send button lets a person pick the model and the depth (Auto, Fast or Deep) for that chat only. It does not change the agent or any other chat. The autonomy control beside it works the same way for the autonomy level.

When to override

I want…	Use
Fast, snappy responses for end users	Fast
The best possible answer for a hard problem	Deep
Lower costs on a high-traffic agent	Fast
A coding assistant that writes production-quality code	Deep
Let the platform decide per task	Auto