📊 Token Usage Dashboard

Workspace context costs, file sizes, and cron estimates. Last scanned: 2026-03-05

9,713

Session Boot Cost

tokens per main session

644

Workspace Files

tracked .md + .json files

799,685

Total File Tokens

across all workspace files

Cron Jobs Active

scheduled cleanup tasks

🚀 Session Boot Cost

Files loaded automatically at the start of every main session with Caelan.

MEMORY.md

~1,874 tok

7.3 KB

AGENTS.md

~1,791 tok

7.0 KB

NIGHTLY_BUILD.md

~1,213 tok

4.7 KB

MEMORY.md

~721 tok

2.8 KB

NIGHTLY_BUILD.md

~716 tok

2.8 KB

AGENTS.md

~644 tok

2.5 KB

SOUL.md

~418 tok

1.6 KB

SOUL.md

~418 tok

1.6 KB

AGENTS.md

~361 tok

1.4 KB

HEARTBEAT.md

~304 tok

1.2 KB

USER.md

~284 tok

1.1 KB

USER.md

~279 tok

1.1 KB

HEARTBEAT.md

~233 tok

0.9 KB

TOOLS.md

~215 tok

0.8 KB

IDENTITY.md

~121 tok

0.5 KB

IDENTITY.md

~121 tok

0.5 KB

Total boot cost~9,713 tokens

Note: Context window is 200k tokens. Boot files use ~4.9% before any conversation starts. MEMORY.md and AGENTS.md are the largest — keep them trimmed.

⏱️ Cron Jobs

Scheduled tasks run in isolated sessions — no memory files loaded.

March 25 Cleanup

Remove March OpenClaw User Group from christchurch-ai.com

0 8 25 3 * (NZST)

~2,000

est. tokens

April 7 Cleanup

Remove April Vibe Coding event from christchurch-ai.com

0 8 7 4 * (NZST)

~2,000

est. tokens

Isolated cron sessions cost ~1,000–3,000 tokens each (prompt only, no memory context loaded). Daily usage tracking is not yet available via OpenClaw API.

📁 Context File Directory

Click a folder to see individual files. Sorted by token cost.

💡 Optimization Notes

✅node_modules/ excluded from git tracking — was ~2M tokens of npm READMEs

✅AGENTS.md trimmed: 2,078 → 352 tokens (removed first-run boilerplate)

✅Stale root files removed: BOOTSTRAP.md, WORKFLOW_AUTO.md, CUSTOMER_AVATARS.md, etc.

✅Duplicate memory/ removed from ai-coaching-collab

⚠️MEMORY.md grows over time — prune old resolved sections monthly into MEMORY_ARCHIVE.md

⚠️Daily usage by LLM not yet available — OpenClaw API doesn't expose token cost per session

💡Use /compact when context hits 75%+ — Puck will alert you via Telegram automatically