Claude Speedometer

A VS Code status-bar indicator showing the token throughput and timing of your latest Claude Code interaction.

The bolt sits in the right-hand status-bar cluster: ↯ 68.20 tok/s, and updates after every interaction. Hover it for a quick stats overlay; click it to open a full stats tab (click again to close).

What it looks like

Hover the bolt for a quick overlay	Click it for the full stats tab

Setup

Once installed, the extension needs Claude Code to export telemetry to it:

1. On first run it offers to configure Claude Code for you. Accept, or run Claude Speedometer: Configure Claude Code Telemetry from the Command Palette. This merges the following into ~/.claude/settings.json (existing settings are preserved):

   {
     "env": {
       "CLAUDE_CODE_ENABLE_TELEMETRY": "1",
       "OTEL_LOGS_EXPORTER": "otlp",
       "OTEL_EXPORTER_OTLP_PROTOCOL": "http/json",
       "OTEL_EXPORTER_OTLP_ENDPOINT": "http://localhost:4318",
       "OTEL_LOGS_EXPORT_INTERVAL": "2000"
     }
   }
   

2. Restart Claude Code so it picks up the new environment.

3. Run a prompt. The bolt updates with the throughput; hover for details.

Reading the stats

The bolt shows the last completed interaction's tok/s. It updates when a turn finishes, not while it is still running, so the figure is stable. The overlay and the tab break the latest turn into sections:

• Speed: output tok/s (output tokens / summed request time).
• Tokens: input, output, cache-write, cache-read. "Output" counts thinking and visible text together; Claude Code's telemetry does not report them separately.
• Timing: total request time and API request count.
• Cost & Model: estimated cost, model, and reasoning effort. Fast mode appears only when it is on.
• Recent: the last several interactions at a glance.
• Context (tab only): the session id and the full workspace path. The session id is the name of the session's ~/.claude/projects/.../<id>.jsonl transcript and the claude --resume <id> handle.

A few things to know about the numbers:

• tok/s is output tokens / summed request time, summing each API call's wall-clock duration_ms (server time plus network and any retries). Tool-execution gaps and time you spend typing or queueing between requests are not counted.
• Output tokens, and therefore tok/s, include both thinking and visible text. Claude Code reports a single output count, so the two cannot be separated here.
• One "interaction" sums all API calls of a single prompt (prompt.id), including tool-call steps and any sub-agents the prompt spawns.
• Stats are global across all Claude Code sessions and windows, not filtered to the current workspace.

Settings

How it works

VS Code extensions can't read Claude Code's internal timing directly, so the extension runs a tiny OTLP/HTTP receiver on localhost:4318 (loopback only), and Claude Code is configured to export its telemetry there. Each claude_code.api_request event carries the token, timing, and cost fields shown above.

It is one machine-wide speedometer: each VS Code window tries to bind the port; the one that succeeds runs the receiver and publishes to a shared file (~/.claude-speedometer/state.json, the last 50 turns / 7 days, safe to delete), and every other window mirrors that file. When the leader window closes, a follower takes over within a couple of seconds.

Development

npm install
npm run compile      # or: npm run watch

To try a change, build a .vsix with vsce and install it:

npx @vscode/vsce package
code --install-extension claude-speedometer-<version>.vsix --force

Then run Developer: Reload Window so the new build loads. The receiver is machine-wide, so reload every open window (or restart VS Code) for the leader to pick up the new code, then restart Claude Code.