First Five Minutes
The fastest product evaluation path: create the plan, select a real device, prove readiness, understand output, and keep artifacts.
Luotsi Docs
Real-device Android automation, replay, and evidence for agents and CI.
Luotsi is the host-side control plane for Android work that has to leave evidence. Start with the first-run path, then choose the workflow that matches the job: operator control, agent exploration, repeatable CI, or replay-driven triage.
Start With The Product Path
Section titled “Start With The Product Path”luotsi quickstart --humanluotsi quickstart --artifacts artifacts/first-run --write-json --write-markdownluotsi doctorluotsi doctor --device <serial> --fixThat sequence gives a human or AI operator the same starting contract: a durable first-run plan, device-selection guidance, readiness checks, and the next command to run.
Installation
Install from GitHub Releases or build from source, then confirm luotsi, version, and devices work locally.
Device Readiness
Use doctor, preflight, lab status, and view setup when the question is whether the attached device is actually usable.
Choose The First Workflow
Section titled “Choose The First Workflow”I need a human/operator loop
Use view when you want a local mirror, hotkeys, screenshots, recording, or remote observation around a real device.
I need an agent loop
Use inspect when an agent needs JSONL snapshots, screen deltas, and structured command results from a physical Android device.
I need repeatable CI
Use scenario-init, scenario-validate, and run when the flow belongs in versioned JSON playbooks with reports and artifacts.
I need to debug later
Use replay and artifacts when the device session is over but the artifact root still needs to explain what happened.
Evaluate Fit
Section titled “Evaluate Fit”First five minutes
Learn the output model first: JSON envelopes, JSONL sessions, artifact roots, replay commands, and the next-action loop.
Device readiness
Use doctor, preflight, and lab-level checks to make device selection and first-run validation explicit before you start debugging the app itself.
Installation
Use the release installer or build from source, then run doctor to confirm adb, device readiness, and live-view prerequisites.
Quickstart
Go from a connected device to a working live-view or scenario run in a few commands.
Live view
Mirror a device to a local SDL window, use presets and profiles, and relay the stream to observer sessions when needed.
Inspect and scenarios
Choose between agent-driven JSONL sessions and repeatable JSON playbooks depending on whether you are exploring or codifying a flow.
AI agent workflows
Use the host-side JSON, JSONL, and replay surfaces Luotsi already exposes when you are building agent loops on top of real Android devices.
Evidence-backed Android Journeys
Combine Android CLI Journey intent with Luotsi scenarios, real-device artifacts, replay, and shared-lab CI evidence.
Agent loop example
Start a minimal host-side adapter that reads inspect JSONL events and writes JSON commands to a real Android device session.
Autonomous discovery
Map visible app states from a real device, preserve agent-readable evidence, and generate review-required scenario candidates.
For engineering leads
Evaluate Luotsi as a product capability: real devices, agent-friendly contracts, CI, shared lab workflows, and artifact trails.
When Luotsi fits
Use the shortest decision page when the question is whether Luotsi matches the workflow shape, team setup, and failure-analysis needs.
Alternatives and comparison
Compare Luotsi with browser automation, hosted device farms, recorder-first tooling, and ad hoc adb scripts.
Reference Catalog
Section titled “Reference Catalog”Command Surface
Lab And CI
Replay And Artifacts
Build And Contribute