Claude Dashboard

- token usage, costs & cheat sheet -

Enter PIN to continue

Headline numbers (API-equivalent cost; you pay $200/mo flat on Max)

Last 30 days (API-equivalent $/day)

Daily breakdown

Date	Tokens	Output	Cache reads	$ equiv.

Sub-agents (in `~/.claude/agents/`, backed up via brain repo)

album-fetcher

Paste a Google Photos share link → returns 25-35 best landscape URLs ready for the JerseyMark "memory" workflow.

build-validator

Curls deployed URLs, checks HTTP/SSL, verifies PIN endpoints. "Is X working?"

disk-investigator

Diagnoses C: drive bloat. Knows about Phone Link / OneDrive Samsung Gallery / Camera Roll triple-mirror trap.

morning-briefing

4-section daily digest (open loops, portfolio, news, today's queue). Also fires from a 5pm cloud Routine.

repo-explorer

"Look at this repo" / "what is X" → tight ~300-word architectural summary.

ship-it

Full deploy dance: git status → selective add → commit → push → poll Vercel → curl-verify the live HTML.

vercel-doctor

Per-project Vercel health: deploys, domains, SSL, env vars present. Read-only.

Cheat sheet (things to type)

Slash commands (in Claude)

/usage: Token usage in this session + history
/cost: $ spent so far in this session
/agents: List installed sub-agents
/status: Model, session length, working dir
/clear: Reset context (cheaper than long sessions)
/resume: Pick a prior session to continue
/init: Bootstrap a CLAUDE.md for this dir
/help: Built-in command list

Token cost (in your terminal)

npx ccusage: Daily token + cost table (this dashboard's source)
npx ccusage monthly: Monthly rollup
npx ccusage session: Per-session breakdown
npx ccusage blocks: 5-hour billing windows (Pro-plan check)
refresh-claude-dashboard.ps1: Update this page (run from anywhere)

MCP & plugins

claude mcp list: Show all MCP servers + status
claude mcp add: Add a new MCP server
claude plugin list: Show installed plugins
claude plugin disable X: Turn off a plugin

Verbal patterns (saves tokens)

"use repo-explorer to...": Force-delegate to a sub-agent
"check Vercel": Triggers vercel-doctor
"check my disk": Triggers disk-investigator
"reset" / "fresh session": Wrap up + start clean
"wrap up" / "end of session": Update brain, commit, push

Model overrides for agents

"check Vercel with Haiku": Override to Haiku for this call (~10× cheaper, snappier)
"check Vercel with Opus": Override upward for harder reasoning
"always use Haiku for X": Permanent: edit ~/.claude/agents/X.md model: line
delete model: line: Agent inherits parent session's model
defaults today: All 7 agents on Sonnet; main session on Opus

Brain locations

SOUL.md, USER.md: Identity (auto-read)
MEMORY.md: Long-term distilled memory (auto-read)
memory/<date>.md: Daily notes
projects/<name>.md: Per-project context
tasks/in-progress/: Active threads
PLAYBOOK.md: Session habits cheat sheet

Useful URLs

personal hub: personal.html
portfolio: portfolio.journeys...
budget: budget.journeys...
admin: admin.html
brain repo: github.com/mfugel/ai-shared-brain