Cline

Cline is an AI coding extension for VSCode and Cursor. It emits structured OpenTelemetry data for every developer task: turns, tokens, cost, cache hits, tool calls, hooks, and a detailed breakdown of code Cline generated versus what the developer accepted or rejected. Routing this to Last9 lets you measure adoption, cost per developer, prompt cache efficiency, and tool reliability — within your existing observability stack.

Cline exports two OpenTelemetry signal types:

• Logs — structured events for prompts, tool invocations, errors, and hook executions
• Metrics — counters and histograms across task, cache, tools, errors, API latency, hooks, AI output, and workspace

What gets exported

Logs (events)

Each Cline session emits structured log records. Filter by service.name = cline (or whatever you set via OTEL_SERVICE_NAME).

Metrics

Cline exports 30+ metric instruments grouped by domain.

Prerequisites

1. Last9 account — Sign up at app.last9.io
2. Cline extension — Installed in VSCode (saoudrizwan.claude-dev) or Cursor
3. OTLP credentials — Get your endpoint and auth header from Integrations → OpenTelemetry

Setup

Cline is configured via environment variables prefixed with CLINE_OTEL_. The variables must be set in the shell that launches your editor — GUI launches via Spotlight, Dock, or Finder will not pick them up.

1. Get your Last9 OTLP credentials

   Navigate to Integrations → OpenTelemetry in your Last9 dashboard. Copy:

   • OTLP Endpoint (e.g., https://otlp.last9.io or https://otlp-aps1.last9.io:443)
   • Authorization header value (e.g., Basic <base64-token>)

2. Add environment variables to your shell profile

   Add the following to ~/.zshrc, ~/.bashrc, or equivalent:

   • HTTP (recommended)
   • gRPC

   export CLINE_OTEL_TELEMETRY_ENABLED=true
   export CLINE_OTEL_METRICS_EXPORTER=otlp
   export CLINE_OTEL_LOGS_EXPORTER=otlp
   export CLINE_OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
   export CLINE_OTEL_EXPORTER_OTLP_ENDPOINT="https://<your-last9-otlp-endpoint>"
   export CLINE_OTEL_EXPORTER_OTLP_HEADERS="Authorization=Basic <your-last9-auth-token>"

3. Reload the shell and launch your editor from the same terminal

   source ~/.zshrc
   cursor .   # or `code .` for VSCode

   Cline picks up the configuration on extension activation.

4. Run a Cline task

   Open the Cline panel and submit any task (e.g. “summarize what this repo does”).

   • Logs flush every 5 seconds
   • Metrics flush every 60 seconds by default

5. Verify data is arriving

   • Metrics — open Metrics in Last9, search for cline_turns_total or cline_tokens_input_total
   • Logs — open Logs, filter by service.name = cline

Configuration reference

Core

OTLP

Per-signal endpoint override

Tuning

What you can do in Last9

Cost and adoption tracking (metrics)

cline.cost.total is a counter you can break down by model and user.email (when set via resource attributes). In Last9 Metrics, plot rate(cline_cost_total[5m]) for spend per minute, or sum over a window for daily totals.

Cache efficiency (metrics)

Compare cline.cache.read.tokens.total against cline.tokens.input.total to measure prompt cache hit ratio. A high ratio means lower per-interaction cost.

Developer outcome (metrics)

cline.ai_output.accepted.* versus cline.ai_output.rejected.* tracks code Cline generated against what developers actually kept. Useful for measuring AI suggestion quality and adoption depth:

sum(rate(cline_ai_output_accepted_lines_added_total[1h]))
  /
sum(rate(cline_ai_output_accepted_lines_added_total[1h]) + rate(cline_ai_output_rejected_lines_added_total[1h]))

Model latency (metrics)

cline.api.ttft.seconds and cline.api.duration.seconds show p95/p99 model performance over time. Plot per model to compare providers.

Tool reliability (metrics)

cline.tool.calls.total and cline.errors.total together highlight which tools fail most. Combine with cline.errors.per_task to spot tasks with elevated error rates.

Hook health (metrics)

cline.hooks.failures.total and cline.hooks.cancellations.total flag hooks that fail or block in flight. Pair with cline.hooks.duration.seconds to find slow hooks.

Team-level tagging

Tag sessions by team via standard OTel resource attributes:

export OTEL_RESOURCE_ATTRIBUTES="deployment.environment=production,team=platform"

All metrics and logs from that session carry team and deployment.environment labels, enabling per-team breakdowns in Last9.

Debugging

Enable detailed Cline OTel diagnostic logging:

export TEL_DEBUG_DIAGNOSTICS=true

Diagnostic output appears in Help → Toggle Developer Tools → Console inside VSCode/Cursor. Includes the resolved configuration, exporters created, connection attempts, and per-export success/failure.

Troubleshooting

• No data in Last9 after running a Cline task

  • Confirm your editor was launched from a terminal that has the CLINE_OTEL_* vars set. GUI launches do not inherit shell env. Run cursor . or code . from a terminal where echo $CLINE_OTEL_TELEMETRY_ENABLED returns true.
  • Check that CLINE_OTEL_TELEMETRY_ENABLED=true (not True, not 1)
  • Wait at least 90 seconds — metrics flush every 60 seconds and may need one cycle to land

• 401 / authentication errors

  • Cline uses comma-separated key=value syntax in CLINE_OTEL_EXPORTER_OTLP_HEADERS, not HTTP header colon syntax. The full string should be Authorization=Basic <token> (with =, not : )
  • Verify the token has not expired in Integrations → OpenTelemetry

• Logs flow but metrics are missing

  • Cline metrics export every 60 seconds — wait at least 90 seconds. Metric names in Last9 use underscores: cline.cost.total becomes cline_cost_total

• Per-signal endpoint customization needed

  • Use CLINE_OTEL_EXPORTER_OTLP_METRICS_ENDPOINT and CLINE_OTEL_EXPORTER_OTLP_LOGS_ENDPOINT to point each signal at a different collector

Please get in touch with us on Discord or Email if you have any questions.