trailblaze

Trailblaze lets coding agents drive real devices and web apps with natural-language steps, then saves sessions as replayable trails for CI tests. It keeps flows repeatable without LLMs at replay time.

Share on XLicense: Apache-2.0

Overview

Trailblaze is a device-automation and testing framework that lets coding agents control real iOS, Android, and web interfaces through natural-language steps. It records each session as a reusable trail, with the intent and actions saved as a file that can be replayed deterministically in CI. This gives teams a way to turn UI interactions into repeatable checks without depending on an LLM during replay.

Key features

  • Natural-language control for real iOS, Android, and web devices
  • Each session saved as a replayable trail file
  • UI snapshot with refs for targeting a visible element
  • CI-ready replay with screenshots, hierarchy, and video reports

Best for

Best for teams that use coding agents to drive live devices or web apps and want a readable record of the flow for deterministic CI replay. It is useful when UI checks need to be repeatable across iOS, Android, and web.

Upstream
block/trailblaze
Fork on GitHub
Guo-astro/trailblaze
Upstream stars
320
Category
AI agents and LLM tools
License
Apache-2.0
Forked
2026-10-02
Sync status
In syncLast synced 2026-10-03