LLxprt Code Telemetry

LLxprt Code never sends telemetry anywhere. There is no network exporter in the codebase — no OTLP, no Google, no vendor endpoint of any kind. Everything described on this page happens entirely on your own machine.

Within that boundary there are two separate layers:

Layer	Default	Where it goes
Session stats	Always on	In memory; shown by /stats
OTEL traces/metrics/logs	Off (telemetry.enabled: false)	A local file, or your console

Session stats are what populate the /stats display — token counts, tool call tallies, and timings for the current session. They are aggregated in memory regardless of the telemetry.enabled setting, are never written to disk by this layer, and disappear when the session ends. Nothing to opt out of: it is your own usage, shown back to you.

The OTEL layer is the one telemetry.enabled controls; it defaults to false, and the rest of this page is about that layer.

This page covers OTEL telemetry (traces, metrics, logs). For conversation logging (the /logging command), see Telemetry Privacy.

What telemetry does

When enabled, the OpenTelemetry SDK registers only HttpInstrumentation — it does not create custom spans for tool calls or model responses. The data emitted is:

• Traces: HTTP request/response spans (auto-instrumented by HttpInstrumentation). There are no custom startSpan calls for tool calls or API interactions.
• Metrics: session counts, tool call counts/latency, API request counts/latency, token usage, file operation counts.
• Logs: configuration events, user prompts (if logPrompts is enabled), tool calls, hook calls, API requests/responses/errors, slash commands.

All data is written locally — to a file if you configure an outfile, or to the console otherwise. The SDK constructs only File*Exporter or Console*Exporter; no OTLP or network exporter is imported anywhere in the source, so there is no code path that could transmit this data.

Enable telemetry

Telemetry is controlled by the telemetry.enabled setting.

Property	Value
Default	false (disabled)
Scope	User settings or workspace settings
Persistence	Saved in settings.json
Precedence	CLI flags override settings files

To enable telemetry, add the following to your user settings or workspace .llxprt/settings.json:

{
  "telemetry": {
    "enabled": true
  }
}

You can also enable it for a single session with the CLI flag:

llxprt --telemetry "your prompt"

CLI flags take precedence over settings files. Specifically, the configuration builder resolves the effective value as argv.telemetry ?? settings.telemetry.enabled — a CLI flag wins if present, otherwise the persisted setting is used. See Configuration for the full settings reference.

Where output goes

Configuration	Output destination
--telemetry-outfile=/path/to/file	All traces, metrics, and logs are written to that file
No outfile configured	Data is written to the console (stdout/stderr)

To write telemetry to a file:

llxprt --telemetry --telemetry-outfile=/tmp/llxprt-telemetry.log "your prompt"

Verify telemetry is working

1. Enable telemetry and start a session.
2. Run any prompt that triggers a tool call or API request.
3. Check the output:
   • File: inspect the outfile with cat or jq.
   • Console: look for OTEL span/log output in the terminal.

Disable telemetry

Telemetry defaults to off, so simply not enabling it is sufficient. If telemetry is enabled in your persisted settings (settings.json), omitting --telemetry on the command line does not disable it — the persisted setting still applies (the builder uses argv.telemetry ?? settings.telemetry.enabled). To explicitly disable telemetry for a single session regardless of persisted settings, use --no-telemetry:

llxprt --no-telemetry "your prompt"

To disable it permanently, remove telemetry.enabled from your settings (or set it to false).

Privacy

• No external transmission: no OTLP or network exporter exists in the source. Data goes only to the configured file or console.
• No data sent to Google or anyone else: there is no upstream endpoint, vendor integration, or opt-in that would transmit telemetry off your machine.
• Session stats are yours: the always-on /stats aggregation stays in memory for the life of the session. It is not persisted and not transmitted.
• Prompt logging: the telemetry.logPrompts setting controls whether user prompt text is included in the llxprt_code.user_prompt log event. Default is true; set it to false to redact prompt content from that specific log event. Note: logPrompts: false does not redact hook input/output data (the llxprt_code.hook_call event always includes hook_input and hook_output fields regardless of this setting).

Logs and metrics reference

The following events and metrics are emitted when telemetry is enabled. All event and metric names use the llxprt_code.* prefix (the service name is llxprt-code, but the telemetry identifiers use underscores):

Logs (event names)

• llxprt_code.config: startup configuration (model, sandbox, approval mode, etc.)
• llxprt_code.user_prompt: user prompt submission (length; prompt text if logPrompts is enabled)
• llxprt_code.tool_call: each tool call (function, args, duration, success)
• llxprt_code.hook_call: hook execution (event name, input, output, duration)
• llxprt_code.api_request: provider API request
• llxprt_code.api_error: provider API error
• llxprt_code.api_response: provider API response (token counts, latency)
• llxprt_code.slash_command: slash command execution
• llxprt_code.next_speaker_check: next-speaker determination
• llxprt_code.conversation_request: conversation API request
• llxprt_code.conversation_response: conversation API response
• llxprt_code.enhanced_conversation_response: enhanced conversation response
• llxprt_code.provider_switch: provider switch event
• llxprt_code.provider_capability: provider capability report
• llxprt_code.tool_output_truncated: tool output truncation event
• llxprt_code.file_operation: file system operation
• llxprt_code.malformed_json_response: malformed JSON from provider
• llxprt_code.model_routing: model routing decision
• llxprt_code.extension_install: extension installed
• llxprt_code.extension_uninstall: extension uninstalled
• llxprt_code.extension_enable: extension enabled
• llxprt_code.extension_disable: extension disabled

Metrics

• llxprt_code.session.count (counter): incremented once per startup
• llxprt_code.tool.call.count (counter): tool call counts
• llxprt_code.tool.call.latency (histogram, ms): tool call latency
• llxprt_code.api.request.count (counter): API request counts
• llxprt_code.api.request.latency (histogram, ms): API request latency
• llxprt_code.token.usage (counter): token usage by type (input, output, etc.)
• llxprt_code.file.operation.count (counter): file operation counts