Telemetry

What StreamView sends to LTC Labs, when, and why. Short version: two anonymous events per session — app_launched and app_closed (with a session_summary blob of counts, feature flags, and config) — plus crash reports if the app blows up, plus an update check at most once a day. No personal data, no terminal content, no commands. Every category is independently opt-out-able in Settings → Data → Privacy (and Updates).

What gets sent today

Per app session — one request

StreamView posts once per session to https://www.ltclabs.dev/api/app/update-check at startup. That request carries:

• app_launched — the new session is starting (sent immediately on init).
• The previous session's app_closed event with its session_summary blob (below) — queued locally on shutdown and flushed at the next startup.

app_closed.session_summary

A single metadata dictionary attached to app_closed. Counts, enums, booleans only — never names, paths, or content.

Field	Type	What it is
session_minutes	int	How long the app was running
streams_opened	int	Streams launched this session
streams_open_at_close	int	Open at shutdown
workspaces_count	int	How many workspaces existed
most_used_layout	enum	Grid / MainSidebar / Rows / Single
used_tail_file	bool	Did you open a Tail File stream?
used_screenshot_paste	bool	Did the screenshot paste fire?
used_ctrl_tab	bool	Did the round-robin shortcut fire?
used_history_search	bool	Did you search history?
custom_icon_uploaded	bool	Any workspace has a custom uploaded image icon?
mcp_enabled	bool	Is MCP integration on?
custom_data_dir_set	bool	Is the data dir pointing somewhere other than the default?
theme	enum	Dark / Light / HighContrast
default_layout	enum	The configured default
default_shell	enum	PowerShell / Cmd / WSL / Git Bash
buffer_size	enum bucket	Scrollback setting (one of four)
idle_timeout	enum bucket	Idle threshold (one of four)
auto_save_sessions	bool	History recording on/off
startup_ms	int	Time from OnStartup to MainWindow shown
memory_mb_at_close	int	GC.GetTotalMemory(false) in MB
template_count	int	How many saved stream templates
pty_process_failures	int	Failed shell launches this session
tokens_today	int	Estimated tokens seen in AI-pane badges today (0 if none). No prompts or names.

Crash reports — only on a crash

Posted to /api/app/crash with the exception type, message, and stack trace. No terminal output, no file paths from your machine, no commands. Gated by Send crash reports.

Update check — at most once per 24 hours

Posted to /api/app/update-check with a random per-machine stagger. Returns the latest version + a support token. Gated by Check for updates automatically.

Support tickets — only when you press Help → Contact Support

Submitted to /api/app/support with what you typed plus the recent app log (no terminal output). You see exactly what's submitted.

What's deliberately not running

• No heartbeat. Earlier builds pinged /api/app/telemetry every 5 minutes (~12 req/hour per running user). Disabled in 2026.5.25 — the timer is configured but never started.
• No per-action events. workspace_added / workspace_deleted / stream_added / stream_closed / layout_changed were removed in 2026.5.25; their useful signal lives in the session_summary blob as aggregates instead.

What is never sent

Not in any payload, not in crash reports, not in update checks:

• Terminal output or scroll-back history
• Commands, environment variables, command-lines, working directories, file paths
• Workspace names, stream names, template names, custom-icon filenames
• Window contents, screenshots, keystrokes
• Your email, account info, or anything you typed (outside a support ticket you compose)

How "anonymous" works

The machine is identified by a SHA-256 hash of Windows' MachineGuid (HKLM\SOFTWARE\Microsoft\Cryptography\MachineGuid). The raw GUID never leaves your machine. There is no user account, no login, and no personal info stored on the LTC side.

The hash gives a stable per-machine identifier so distinct machines (DAU / retention) can be counted without revealing whose machine. Different machines = different hashes; reinstalling Windows = a new hash.

Why we send anything at all

• Is anyone using it? Counting distinct hashed machine IDs per day = daily / weekly / monthly active users.
• Do they come back? First-seen vs last-seen of each hash = retention.
• Does the app stay up? Crash reports + pty_process_failures surface regressions across releases.
• Are we shipping the right defaults? most_used_layout, theme, default_shell, etc. tell us when the shipped default is wrong (e.g. if everyone changes Theme to Dark, Dark should be default).
• Which features matter? The used_* flags tell us which features earn their keep so we know where to invest next.

Opt-out

Three independent toggles, all on by default, all in Settings:

• Send anonymous usage data (Settings → Data → Privacy) — when OFF, suppresses app_launched, app_closed, and the session_summary blob, and drops any already-queued usage events without transmitting them. The startup /api/app/update-check request still happens (it's how the app gets its support token and license status), but its Telemetry payload is null.
• Send crash reports (Settings → Data → Privacy) — when OFF, suppresses /api/app/crash.
• Check for updates automatically (Settings → General → Updates) — when OFF, suppresses the periodic update check.

Turning all three off leaves only the startup /api/app/update-check request (no event payload) and the support-ticket endpoint (only fires when you submit a ticket).

Related

• FAQ — Does StreamView send my terminal output anywhere?
• Settings

---

Developer notes

Internal implementation reference. The canonical contract lives in ltc-website/srcDotNet/docs/TELEMETRY_CONTRACT.md; this section is the StreamView-specific wiring.

App identifier in payloads: streamview

Wire model (current)

When	Endpoint
Once at startup	POST /api/app/update-check — carries app_launched and any queued events (including the previous session's app_closed + session_summary)
On unhandled exception (gated)	POST /api/app/crash
At most once per 24h, gated	POST /api/app/update-check (update-check path)
When the user submits a support ticket	POST /api/app/support

No heartbeat. Expect ~1 POST per app session per machine, plus rare crash / update / support requests.

Event vocabulary (current implementation)

Event	Status
app_launched	✓ Required. Plain (no metadata).
app_closed	✓ Required. Carries the session_summary metadata blob built by LtcLabsService.EmitAppClosed.
layout_changed, stream_added, stream_closed, workspace_added, workspace_deleted	— Removed in 2026.5.25 (replaced by session_summary aggregates)
probe_nested, probe_direct	— Removed (internal smoke-test events)

Implementation checklist

• Trim event vocabulary to app_launched + app_closed (2026.5.25)
• Stop the 5-minute heartbeat (timer remains configured but never started; 2026.5.25)
• Implement app_closed.session_summary (2026.5.25): counters in LtcLabsService.NoteXxx, snapshot built in EmitAppClosed, MainWindow's OnClosed fills in workspace/stream/custom-icon extras, App's OnStartup feeds startup_ms via a Stopwatch.
• Settings → Data → Privacy: Send anonymous usage data + Send crash reports. Check for updates automatically lives under General → Updates. (2026.5.25)
• On-close flush is batched — events queue to disk via SaveQueuedEvents and flush on next InitAsync (one POST).

File locations

Thing	Path
Event queue (offline buffer)	%APPDATA%\StreamView\telemetry_queue.json
Last update-check timestamp	%LOCALAPPDATA%\StreamView\.last-update-check
Logs	%LOCALAPPDATA%\StreamView\streamview.log

About-dialog disclosure

Standard paragraph from the LTC contract:

StreamView sends diagnostic and service-improvement data to LTC Labs to help us detect and fix bugs, improve stability, and understand which features are working. This data is anonymous — no personal files, content, names, or account identifiers are transmitted. You can opt out of optional usage data in Settings → Data → Privacy. Crash reports are togglable separately so you can keep stability reporting on while opting out of usage data.