Amber deer markAmber Team Case Study GitHub
Amber deer mark

Amber

A self-hosted runtime that lets agents resume work after failures and makes debugging agent runs easier.

Read the Case Study → Download on GitHub

01 The code

Crash recovery is a few lines away.

Import, annotate, run.

Turn agent code into workflows that resume after failures with a few decorators.

Amber SDK code snippet showing crash recovery decorators

02 Durability

Failures no longer erase progress.

Amber preserves your agent’s work so it never starts over.

Without Amber an agent restarts from the beginning after a crash; with Amber it resumes from the last saved step.

03 Observability

Agent-specific metadata is captured at each step.

When something fails, inspect and debug workflows.

Amber dashboard inspector showing a recovered Research Assistant run

How we built it.

Read the Case Study →