Wipe the Ads Off Your TV Home Screen, Plus 4 More Free AI Tools

Someone vibe coded their TV a new home screen with zero ads, in one afternoon, with Claude Code. The repo, CleanYourTV (it was called tv-vibehack until 19 October), is free, and it's one of five free AI tools on GitHub worth installing this week.

The other four: Orca runs Claude Code, Codex and other coding agents side by side. Indra turns scanned PDFs into text on an ordinary CPU. Scout is a browser agent that clicks through boring web tasks for you. And earshot lets Claude hear.

Four of the five were created between 13 and 17 October 2026. Here's what each does, the catches, and how to install it. Star counts are from GitHub on 26 October.

What do these free AI tools on GitHub do?

CleanYourTV: a TV home screen with zero ads

CleanYourTV switches off the preloaded bloat, ad services and sponsored rows on an Android TV or Google TV over Wi-Fi, then installs a clean home screen that looks like an Apple TV. Paste its prompt into Claude Code, Codex or Gemini CLI and the assistant does the work, asking before it touches anything you use. It was created on 17 October and had over 37,000 stars by 26 October.

Nothing is uninstalled, and a restore script puts everything back. It doesn't work on Samsung or LG, Fire TV gets the clean-up but not the new home screen, and apps like YouTube still show their own ads.

Orca: run your coding agents side by side

Orca is a free desktop app for running Claude Code, Codex, OpenCode and other coding agents side by side, each in its own copy of your project, all tracked in one place. Send one prompt to several agents, compare what comes back and keep the best. A phone app lets you check on them from anywhere.

It runs on your own agent subscriptions and has over 106,000 stars. On 19 October alone, it picked up 1,146.

Indra OCR: scanned PDFs to text, no graphics card

Indra turns PDFs and images into Markdown, plain text or JSON, on an ordinary CPU. It reads more than 80 languages, even on pages scanned sideways. Its makers clock it at about 44 pages a second on a 32-core CPU, around 15 times faster than Tesseract. On 4 cores, it's about a fifth of a second a page.

The code is Apache-2.0, but the model is free only for research, personal use and startups under $20M in funding or revenue. For client work, check Datalab's licence first.

Scout: hand a boring web task to a browser agent

Scout is an open-source agent that uses a real browser to finish a task you describe in plain words, like comparing three prices. It reads the page, clicks, types and scrolls until the job's done. If it can't finish, it tells you what's missing and returns what it found.

It works with almost any AI model, including one on your own computer, and runs best on Outcrawl's cloud browser, which gives you 1,000 free credits a month with no card.

earshot: let Claude hear your audio

earshot gives Claude Code and other coding agents ears. It describes a sound in words, finds the moments in a long recording that match a sound you describe, and scores a video's soundtrack against seven professional launch films, with timestamped fixes. It runs locally, on your CPU.

It was created on 15 October and has over 2,800 stars. Its README is upfront that it can't judge taste, emotion or what's being said.

How do you install each one, and what should you try first?

CleanYourTV. On the TV, go to Settings → Device Preferences → About and press OK on Build seven times. Then, in Developer options, turn on USB debugging (and Network or ADB debugging, if it's there). On a computer on the same Wi-Fi:

git clone https://github.com/Gathin23/cleanyourtv

Open that folder in Claude Code, paste in the repo's PROMPT.md with your TV model filled in, and answer its questions. It takes about fifteen minutes. Only do it on a TV you own: changing its software can affect the warranty.

Orca. Download it for Mac, Windows or Linux from onorca.dev. On a Mac with Homebrew:

brew install --cask stablyai/orca/orca

Then send the same prompt to two agents and keep the better result.

Indra. You need Python 3.10 or newer. The first run downloads about 76 MB of models.

pip install indra-ocr
indra-ocr paper.pdf

Swap paper.pdf for your scan, and you get Markdown you can paste straight into ChatGPT or Claude.

Scout. You need Node 22 or later.

npm i @outcrawl/scout

Add a free Outcrawl key and a key for your model, then follow the three-step quickstart in its README. Not writing code yourself? Hand that README to your coding agent and ask it to set Scout up and compare three prices for you.

earshot. You need Python 3.10 or newer and ffmpeg. On Windows, use .venv/Scripts/python in place of .venv/bin/python.

git clone https://github.com/Leonxlnx/earshot && cd earshot
python -m venv .venv
.venv/bin/python -m pip install torch --index-url https://download.pytorch.org/whl/cpu
.venv/bin/python -m pip install -e ".[mcp,llm,dev]"
.venv/bin/python -m earshot doctor --load

Then add it to Claude Code, with the path to your earshot folder:

claude mcp add earshot --scope user -e EARSHOT_THREADS=3 -- /path/to/earshot/.venv/bin/python -m earshot mcp

The first time Claude uses it, it downloads about 1.1 GB of open models. Then give Claude a video and ask it to critique the sound with earshot.

Running more than one agent at a time? See our GitHub repos for Claude Code. To find the jobs costing your team the most time, take the free AI bottleneck test, or read our insights on AI implementation.