agentmon reads your local Claude Code, Codex & OpenCode transcripts and turns them into tokens, cost, models, sessions and rhythm — all on your machine. Nothing is uploaded.
…or drag & drop your .claude / .codex folder here
.claude and .codex are hidden folders, so the file dialog won't show them by default. Easiest fixes:
1. Jump straight to it. In the dialog press the dialog’s path field, paste a path below, and open it — this reveals the hidden folder in one step. Picking .claude directly works.
2. Or reveal hidden files (enable “show hidden files” in the dialog), then open the folder directly.
3. Or pick your home folder — nothing else on your disk is read; we only step into .claude / .codex.
Each pick connects one folder — connected one tool but not the other? Use “add folder” in the header to merge more in.
A polished web dashboard or a keyboard-driven terminal — same numbers, rendered live from a demo dataset below.
Every token in and out, cache included, with estimated spend per model — charted over time.
Hit-rate and the dollars caching actually saved you over time.
Which model did the work — usage, share and cost, side by side.
Ranked by usage; drill into any session's timeline, tools and models.
Top tool calls, web search/fetch, and the main-vs-subagent token split.
An hour × weekday heatmap of when you actually ship code — explore it live in the showcase below.
Open it in any Chromium browser, point it at your folder. Files are read on-device — nothing is uploaded.
Clone the repo and run the Next.js app against your real ~/.claude & ~/.codex for the complete dashboard.
A fast, keyboard-driven TUI right in your terminal — no install needed. Mouse + charts included.
Every way of running agentmon reads your transcripts on your machine and aggregates them locally. There is no upload, no account, and no analytics back-end — your data never leaves your device.
No account, no upload, no setup — run npx agentmon, or connect your data right here in the browser.