Track LLM token spending, coding agent runs, model benchmarks, and prompt experiments on your desktop wallpaper. No dashboards to open. No tabs to remember. Glance down and know your AI budget.
See daily/weekly API spend without opening billing dashboards. Inline commands poll your provider's usage API and display costs in real-time cells.
Watch Claude Code, DeepSeek-TUI, or custom agents complete tasks. Inline process cells show live stdout from your agent orchestrators.
Keep a persistent comparison table — latency, cost/1K tokens, quality scores — visible while you code. Update it as you benchmark new models.
Draft prompts in cells, run them with r: prefix commands, see results in adjacent cells. Iterate without context-switching.
No Python, no Node, no Docker. A single .NET binary. No supply chain to audit. Runs on the same Linux box as your GPU cluster.
All data stored as CSV — pipe it, grep it, git-diff it, feed it back to your LLMs. No proprietary formats. Export to SQL, Markdown, HTML, or JSON.
A typical QuickSheet layout for an AI engineer working with multiple LLM providers:
| A | B | C | D | E | |
|---|---|---|---|---|---|
| 0 | 🤖 AI Ops | Provider | Today $ | MTD $ | Budget % |
| 1 | i: curl -s $ANTHROPIC_URL/usage | Anthropic | $4.82 | $127.40 | 63% |
| 2 | i: curl -s $OPENAI_URL/usage | OpenAI | $1.20 | $34.50 | 17% |
| 3 | i: ./check-deepseek-balance.sh | DeepSeek | $0.31 | $8.90 | 4% |
| 4 | Total | Σ | Σ | ||
| 5 | |||||
| 6 | 🏃 Agent Runs | Agent | Status | Tokens | Duration |
| 7 | i: gh run list --limit 1 | Claude Code | ✅ Done | 48,230 | 3m 42s |
| 8 | i: ./agent-status.sh deepseek | DeepSeek-TUI | 🔄 Running | 12,891 | 1m 15s |
| 9 | i: ./agent-status.sh local | Ollama Local | ⏸ Idle | — | — |
Query local Ollama models directly from cells. Summarize text, generate code, translate — results appear in adjacent cells. Zero cloud dependency.
ext: ollama ask "Explain this error"
Run any shell command and display live output. Poll APIs, tail logs, check agent status — output refreshes automatically in the cell.
i: curl -s api.anthropic.com/v1/usage | jq .today
One-shot commands triggered on demand. Run benchmarks, restart agents, deploy models — all from a cell click.
r: python bench.py --model gpt-4o-mini
Automatic sum of numeric columns. Perfect for totaling daily token costs across providers or aggregating benchmark scores.
Row 4: Σ → auto-sums API costs above
GitHub Copilot integration. Ask questions, get code suggestions, summarize PRs — all from within your wallpaper spreadsheet.
ext: copilot "review this diff"
Dedicated AI cost tracker. Built-in pricing for 15 models (GPT-4o, Claude, Gemini, DeepSeek). Budget alerts, provider comparison, monthly spend log — all in your wallpaper.
ext: ai-costs usage --provider openai --days 7
Export your cost tracking data to Markdown, HTML, JSON, or SQL. Feed structured data back into your AI pipelines or documentation.
dotnet run -- costs.csv --export-sql costs.sql
Poll each provider's usage API every 5 minutes. Column sums give you real-time daily totals. Cells turn red when approaching budget limits.
i: curl -sH "Authorization: Bearer $KEY" https://api.anthropic.com/v1/usage/daily | jq '.cost_usd'
Monitor multiple concurrent coding agents. See which are running, idle, or failed. Live token counts and durations update in real-time.
i: gh run list --repo my/project --workflow agent.yml --limit 3 --json status,conclusion -q '.[0].status'
Keep a grid of prompt variants with quality scores. Run each prompt via r: prefix, paste results, score them. CSV export feeds your evaluation pipeline.
r: echo "Prompt v3" | python eval_prompt.py --model claude-sonnet --metric coherence
Persistent comparison of models you're evaluating. Latency, cost, quality, context window — all visible while you code.
i: python benchmark.py --models "gpt-4o,claude-sonnet,deepseek-v4" --format csv | tail -1
Inline command checks if daily spend exceeds threshold. Combine with r: to auto-pause expensive agents when budget is blown.
i: [ $(curl -s $USAGE_API | jq .today_usd) -gt 50 ] && echo "⚠️ OVER BUDGET" || echo "✅ $32.40"
Track fine-tuning jobs across providers. See training loss, completion percentage, and ETA without leaving your IDE.
i: openai api fine_tuning.jobs.list --limit 1 | jq '{status,trained_tokens,model}'
A purpose-built extension for AI API spend. Built-in pricing for 15 models. No browser dashboards, no SaaS — just your wallpaper showing exactly where your money goes.
Connects to OpenAI /v1/usage API. Shows daily spend, request count, and token breakdown by model — refreshing on your wallpaper.
ext: ai-costs usage --provider openai --days 7
Instant cost estimates with shorthand notation. Supports input/output token splits. Covers GPT-4o, Claude Sonnet/Opus, Gemini Pro, DeepSeek V3.
ext: ai-costs calc gpt-4o 100k 5k
Side-by-side pricing table for all 15 models. Sort by input cost, output cost, or total. Find the cheapest model for your workload.
ext: ai-costs compare --sort total
Set monthly limits per provider. See progress bars, projected end-of-month spend, and ⚠️ alerts when approaching budget caps.
ext: ai-costs budget --limit 200 --provider anthropic
gh repo clone cemheren/quicksheet-ai-costs cd quicksheet-ai-costs dotnet build -c Release
| Feature | QuickSheet | LangSmith / Helicone | Custom Grafana | Browser Tabs |
|---|---|---|---|---|
| Always visible (no switching) | ✓ Desktop wallpaper | ✗ Browser tab | ✗ Browser tab | ✗ Must switch |
| Zero setup / dependencies | ✓ Single binary | ✗ SaaS signup | ✗ Docker + config | ✓ Just open |
| Custom scripts / commands | ✓ i: and r: prefix | ✗ Fixed UI | ~ Plugins needed | ✗ Manual copy |
| Multi-provider view | ✓ Any API via curl | ~ Provider-specific | ✓ With datasources | ✗ Multiple logins |
| Data export (CSV/SQL/JSON) | ✓ Built-in | ~ API only | ~ CSV plugin | ✗ Manual |
| Offline / air-gapped | ✓ Local-first | ✗ Cloud required | ✓ Self-hosted | ✗ Online only |
| Cost | Free & open source | ~ Free tier limited | Free (self-host) | Free |
# Clone and run git clone https://github.com/cemheren/QuickSheet cd QuickSheet dotnet run -- examples/ai-workflow.csv --desktop # Or use a pre-built release gh release download --repo cemheren/QuickSheet -p '*.zip'
Requires .NET 9 SDK. Linux (X11) and Windows supported. Zero NuGet dependencies.
Stop switching tabs to check token costs. Put your AI engineering dashboard where you'll actually see it — on your desktop.
⭐ Star on GitHub Browse Extensions