Skeptacular Apps
Free · Open source · Local-first

Tools that run on your machine.

Free, open-source tools that run locally. Small, honest utilities by Matt Skepner, with no telemetry and no backend.

The apps

Six small tools, one point of view.

Each one does one job well and keeps your data on your device. None has shipped yet; here is where each one stands.

Active Clipboard

The clipboard manager for people who live in AI tools: local-first and secret-safe. Searchable history, pins, snippets, and context packs that bundle several clips into one paste. Detected secrets are masked and expire after 2 minutes.

Mac menu barIn development

Token Usage Monitor

TUM

One glance at every AI limit before you hit it. Segmented meter bars for your AI plans and spend: Grok, Claude, ChatGPT/Codex, Cursor, Gemini, and more. Your keys stay on your Mac.

Mac menu bar (macOS 14+)In development

Sniffbox

A local scanner that checks agent skills, MCP configs, and repos for prompt injection and leaked secrets before you trust them. It never runs tools based on what it reads.

Command lineComing soon

Owltally

A local-only dashboard that shows what your AI coding agents did while you were away, with a shareable “my agent’s week” card. Bound to 127.0.0.1.

Local dashboardComing soon

ModelDerby

Race local LLM setups against each other. Result cards print the hardware, model, quant, tokens per second, and method, so every number is reproducible. Local models by default.

Local benchmark toolIn development

Active Clipboard Snippets

Keyboard

Your snippets as a phone keyboard. Unlimited plain snippets, free. Everything stays on your phone, and the keyboard is designed never to ask for network access.

iPhone and AndroidComing soon
Latest

ModelDerby ran its first race.

Oct 18, 2026. This is the real results card it produced, not a mockup. More in the build log.

ModelDerby results card. Prompt: Name three primary colors in one short sentence. 1st: creative, temperature 1.0, time to first token 0.09s, total 0.28s, 69.8 tokens per second. 2nd: precise, temperature 0.2, 0.13s, 0.34s, 60.6 tokens per second. Raced locally. No cloud, no telemetry.
Real output from ModelDerby’s first race, run locally on a DGX Spark on Oct 18, 2026: two settings of the same local model. Temperature 1.0 beat 0.2 (0.28s vs 0.34s total, about 70 vs 61 tok/s). One prompt, one run, not a benchmark claim.

Free and open source

Every app is free, and it stays free. The code will be public on GitHub.

Runs locally

No backend to trust. Your clips, keys, and logs stay on your machine.

No telemetry

The apps don’t phone home. See the privacy page.

Free, and it stays free.

Follow along on GitHub, where the repos will appear as they go public. If an app saves you time later, you can tip what it’s worth to you.