Greynewell Tokentrace
greynewell-tokentrace is a Claude Code plugin for the Operate phase that traces token usage with spans, latency percentiles, and alerts.
by greynewell · github.com/greynewell/tokentrace
Install when agent sessions burn tokens or feel slow and you need spans, latency percentiles, and alerts to see where usage actually goes.
Add it to Claude Code
Install the plugin in Claude Code. One command, paste-ready.
/plugin install greynewell-tokentrace@greynewell/tokentraceBuilt to be called by your agent
Skillselion is itself an MCP server. Your agent can pull this entry and a paste-ready install config straight from the API - no copy-paste.
Retrieve this entry with skillselion.get_details("plugin:greynewell/tokentrace") and the paste-ready config with skillselion.get_install_config("plugin:greynewell/tokentrace").
What it does
greynewell-tokentrace is a Claude Code plugin marketplace entry that helps solo builders answer a blunt question: where did your tokens go? Instead of guessing from bill shock, you get span-oriented tracing, latency percentiles, and alerts suited to agent sessions that fan out across tools and sub-calls. Indie developers running Claude Code daily benefit when a feature ship suddenly doubles spend or when responses lag without an obvious bottleneck. The plugin sits primarily in Operate monitoring, but the same metrics support Ship performance tuning and Grow analytics when you correlate usage with experiments. It is an observability integration—not a code generator—so expect to install it once agents are in real use, wire alerts you will act on, and review traces after heavy debugging or automation recipes. Community metadata points at enrichment and recipe-oriented tooling; treat it as operational instrumentation layered on top of your existing skills rather than a replacement for tests or security review.
Highlights
- Answers where did your tokens go with trace-style visibility into agent runs
- Latency percentiles for spotting slow tool or model paths
- Alerting hooks when usage or performance crosses thresholds
- Chef/recipe-style tooling keywords suggest composable observability recipes for Claude workflows
- Single-plugin bundle focused on cost and performance forensics
Why builders use it
Agent bills and slow sessions feel opaque when you cannot see which spans, tools, or steps consume tokens and time.
After install, you can inspect traces and percentiles and get alerts so you trim wasteful agent paths and fix latency hotspots.
At a glance
- Type - Plugin in DevOps.
- Adoption - 0 installs, 5 stars, 0 votes.
FAQ
Who is greynewell-tokentrace for?
It is for Claude Code users—especially indie builders—who run agents often and need to understand token consumption and latency at span level.
When should I use greynewell-tokentrace?
Use it in Operate when costs climb or sessions slow down, and during Ship perf work when you want percentile-based latency evidence before scaling usage.
How do I add greynewell/tokentrace to my agent?
Install the greynewell/tokentrace plugin from its marketplace repo in Claude Code, configure tracing and alerts per the project docs, then run agent tasks you want measured.
Comments
Share how you use greynewell-tokentrace, gotchas, or tips for other indie builders.
No comments yet - be the first to share how you use it.