Questions and answers.
Everything you need to know about Ansight.
What is Ansight?
Ansight is an SDK and local CLI host that gives developers and AI agents runtime visibility into mobile apps. The SDK captures evidence inside the app, structured ansight commands let terminal-capable agents inspect sessions and call guarded tools, and the local host player provides visual review in a browser.
How do I use Ansight?
Ansight provides a natural language bridge to the internals of your app. We expose discrete, primitive operations that any AI agent can build on. You could ask your agent to inspect a database by running a remote SQL query, analyze the current visual tree, take a screenshot, or download a sandboxed file from your app. We provide the tools; your creativity defines how you use them.
Which mobile frameworks are supported?
.NET MAUI, native iOS (SwiftUI/UIKit), native Android (Jetpack Compose), React Native and Expo development builds, Flutter, and the Cordova-family Capacitor bridge are in beta, with framework-aware widget or DOM inspection alongside native visual trees.
How does CLI-based agent access work?
The resident Ansight host receives evidence from paired apps. Agents run versioned ansight commands with --json to request screenshots, logs, telemetry, sessions, and app-defined tools. The agent gets structured runtime evidence without a separate client bridge.
Does my data leave my machine?
The Ansight host runs locally and the SDK communicates directly with it over your local network; Ansight does not route telemetry, sessions, or app data through a cloud service. The current low-friction developer protocol uses clear-text UDP and WebSocket traffic, so use it only on a trusted development or QA network.
Can I use Ansight in debug and release builds, and on simulators or physical devices?
Yes, but remote tooling and enrollment should remain in trusted development or QA variants. Ansight works on simulators, emulators, and physical devices. An enrolled phone or tablet can stream app/session logs, SDK diagnostics, telemetry, screenshots, touches, and exposed runtime tools; use the platform tooling separately for complete device and OS console logs.
How is pairing secured?
Enrollment is intentionally simple: the CLI displays a short-lived, one-use QR invite; the app scans it, registers its stable app-private installation ID, and stores the returned access token for later reconnects. This prevents casual reuse of the invite, but the current clear-text protocol does not encrypt traffic or authenticate the network path. Treat the QR as a bearer credential and use Ansight only on a trusted local network.
What kinds of things could I use Ansight for?
Ansight is useful anywhere runtime evidence matters. Common examples include extracting high-quality bug reports, creating flow chart diagrams from a user's test session, analyzing memory usage, investigating trend regressions, understanding reproduction paths, and mapping user flows from real captured behavior instead of guesswork.
What AI agents work with Ansight?
Any coding agent that can run local shell commands can use Ansight. Codex, Claude Code, Cursor, CI jobs, and ordinary scripts all call the same ansight command surface and consume versioned JSON output.
Is the CLI the primary interface?
Yes. The Ansight CLI is the canonical interface for host lifecycle, pairing, sessions, app tools, devices, tests, tasks, triggers, automation, and CI. The resident host also serves a local browser player for interactive session review.
How does the QA tester workflow work?
The SDK continuously captures logs, metrics, screenshots, and navigation state during QA. When a tester hits a bug, the host preserves the session and the local player shows its timeline, evidence, and user journey. The tester can export or share the session with a developer, complete with reproduction steps and supporting evidence.
What kind of analysis does the AI provide?
The AI agent reconstructs the execution timeline from the captured session, identifies anomalies, and correlates code paths with runtime behavior. It provides root-cause guidance with specific file and method references — not generic suggestions, but analysis grounded in real execution data.
Can I choose what the AI agent sees?
Yes. Selective analysis inputs let you decide exactly which data the agent receives: logs, FPS metrics, memory samples, screenshots, device metadata, and linked app context. You stay in control of the investigation.
How do I pair a device?
Run ansight pairing issue <app-id> --qr, or start a host with ansight host run --pair <app-id>, then scan the terminal QR from the app's developer-only scanner. The device connects over your trusted local network.
Is the SDK open source?
Yes. The Ansight SDK is open source and available on GitHub. The Ansight CLI package includes the resident host and local browser player.
What does it cost?
Ansight is currently in early access. Download the current build to try the beta SDKs, including Flutter and Cordova/Capacitor — pricing will be announced closer to launch.
Can I export sessions?
Yes. You can package any session into a portable archive and attach it to a ticket, hand it to another developer, or revisit it later. Saved analyses can also be browsed, filtered, and reopened by app and session.
Ansight turns mobile apps into a structured tool surface for agents.
Install the Ansight CLI and connect a supported mobile app to the local host.