How the data works
Three sources, all local. Nothing is sent to tomorrowtools, ever — there's no server on our end to send it to.
Claude Code exports it locally
Session transcripts, OpenTelemetry metrics, hook events, and the statusLine payload — all already produced by Claude Code itself, on your disk, as part of how it normally runs.
tomorrowtools reads them
A small local process tails your session logs and receives the OTel/hook data over localhost — never over the network, never off your machine.
You see it, live
A dashboard on localhost updates every few seconds as new data arrives. Close it, delete it, audit the source — it's entirely yours.
The three data sources, in detail
OpenTelemetry export
Claude Code can export token usage, tool decisions, and skill/agent attribution as standard OTel metrics. tomorrowtools receives these directly over a local HTTP endpoint — the same mechanism you'd point at any OTel collector, just pointed at localhost instead.
Hooks
SessionStart, SessionEnd, SubagentStart, and SubagentStop hooks fire on real session/agent boundaries. tomorrowtools captures these to build the session → task → agent hierarchy you see in the dashboard.
statusLine
Claude Code's statusLine payload carries your real-time rate-limit percentages — the same numbers /usage shows — forwarded to the dashboard on every turn.
What never happens
No account creation. No API keys sent anywhere but your own machine. No telemetry back to us. No cloud dependency of any kind — if your machine has no internet connection, the dashboard still works, because it never needed one.