Command Reference

tauri-agent-tools provides 38 commands for inspecting, interacting with, monitoring, post-mortem analysis, and diagnosing Tauri applications. Inspection commands are read-only. Interaction commands are debug-only (require the dev bridge). Bridge-free diagnostics work on dead apps and release builds.

Command Summary

Categories

Visual Capture

• screenshot — capture full windows or specific DOM elements with real screen pixels
• diff — compare two screenshots with pixel-level difference metrics
• snapshot — capture screenshot + DOM tree + page state + storage in a single call

DOM Inspection

• dom — explore DOM tree structure, accessibility tree, computed styles
• eval — run arbitrary JS expressions in the webview (supports --file for loading from files)

Monitoring

• ipc-monitor — watch Tauri IPC calls with timing and filtering
• console-monitor — capture console.log/warn/error output
• rust-logs — monitor Rust tracing/log output and sidecar processes
• mutations — watch DOM mutations with attribute tracking

State Inspection

• storage — read localStorage, sessionStorage, and cookies
• page-state — URL, title, viewport, scroll position, Tauri detection
• store-inspect — inspect reactive store state (Pinia, Vue devtools, custom hooks)

Window Management

• info — window geometry, position, display server
• list-windows — enumerate windows with Tauri detection
• wait — poll for windows, DOM elements, or JS conditions

Interaction (debug-only)

• click — click, double-click, or right-click a DOM element
• type — type text into an input field (supports --clear)
• scroll — scroll by pixels, to top/bottom, or scroll an element into view
• focus — focus a DOM element
• navigate — navigate to a route or URL
• select — select a dropdown value or toggle a checkbox
• invoke — call a Tauri IPC command with JSON payload

Workflow

• probe — discover running bridges, check health, list window labels; when no bridge is discoverable and no explicit --port/--token/--pid is given, emits a complete result with target.alive: false and an actionable note instead of erroring
• capture — collect screenshot + DOM + page state + storage + console errors + Rust logs into a bundle
• check — run structured assertions against DOM state (selector exists, text matches, no console errors)

Bridge-free diagnostics (new in 0.7)

Work without the dev bridge — for release builds, dead apps, and sidecar processes.

• app-paths — resolve a Tauri 2 app's OS data/log/cache/config directories from tauri.conf.json
• config inspect — emit a structured snapshot of tauri.conf.json plus a capability/permission audit
• os-logs — tail the host OS log stream filtered to a Tauri bundle id (NDJSON envelopes)
• sidecar tap — wrap-and-run a sidecar binary, frame its stdout as NDJSON, validate against an optional JSON Schema
• sidecar replay — replay a recorded NDJSON stream to stdout or into a fresh process; --tap-format unwraps {dir,ts,line} tap wrapper rows, optionally filtered by --dir in|out
• forensics — one-shot bundle for post-crash analysis (composes the above; works on dead apps)
• logs — merge an app's scattered logs (on-disk tauri-plugin-log files + the live bridge /logs ring buffer) into one timestamp-ordered NDJSON stream, normalized to UTC; filter by --level/--source/--filter and --correlate to infer correlation ids (works with no bridge); --follow tails the live bridge via v0.8 non-draining cursor reads, falling back to drain-polling on older bridges (--interval <ms> sets the fallback poll interval, default 2000)

Bridge-extending diagnostics (new in 0.7, requires bridge v0.7.0+)

Talk to new dev-bridge endpoints. Each feature-detects via GET /version and surfaces a clear upgrade error against older bridges.

• process-tree — Tauri PID + registered sidecars rendered as a tree
• capabilities audit — live audit of declared Tauri capabilities, flags wildcard and over-broad scopes
• webview attach — webview inspector URL or platform hint (webview2, webkitgtk, wkwebview)
• health — uptime + webview readiness + per-sidecar liveness; exits non-zero when unhealthy

Super-command

• diagnose — best-effort. Composes forensics with live bridge data when reachable; degrades cleanly when not. Master summary.md includes a "Next steps" section pointing at the right deeper command.
• bundle — collect a shareable incident archive (merged logs + deep process-tree + app-paths + forensics, plus an optional UI capture) into one directory and a .tar.gz. Secrets (tokens, API keys, passwords, JWTs, AWS keys) and PII (emails, IPs, phone numbers, home paths — including base64-encoded ones) are redacted from text artifacts on write; JSON/NDJSON artifacts are re-serialized so they stay parseable; unredacted images are flagged as warnings; each phase degrades cleanly.

Common Patterns

JSON output

All commands that produce structured output support --json:

tauri-agent-tools info --title "My App" --json
tauri-agent-tools dom --json
tauri-agent-tools storage --json

Bridge auto-discovery

Commands that need the bridge automatically discover it via token files in /tmp/. You can override with explicit flags:

tauri-agent-tools dom --port 9876 --token abc123

Duration-based monitoring

Monitor commands accept --duration to auto-stop:

tauri-agent-tools ipc-monitor --duration 5000
tauri-agent-tools console-monitor --duration 10000 --level error
tauri-agent-tools rust-logs --duration 10000 --level warn