Live now· #62 this week
Code Scout

Code Scout

Free native desktop AI coding agent that runs local models

FreeAICodeAI Coding37 impressions#62 this week1 comment+1 today

Comments

>log in to comment
  • Frank Vitetta[maker]· 2d ago

    Hi everyone, Frank here, maker of Code Scout. I built it because AI coding tools got expensive and most of them route your repo through someone else's cloud. Code Scout is a native desktop agent that runs local models (Ollama, LM Studio, llama.cpp) or cloud ones if you choose, with shell, file and git tools. Free for personal use. Happy to answer questions about the architecture or local model setups.

Code Scout is a free native desktop AI coding agent built by Frank Vitetta. It runs local models through Ollama, LM Studio or llama.cpp, or cloud providers if you prefer, and works directly in your repository with shell, file and git tools. It plans a task, executes it step by step with real tools, and iterates on the result using your chosen model. Your code stays on your machine unless you pick a cloud model. Free for personal use on macOS, Windows and Linux.

More AI tools →80 free dev tools →

  • Free desktop AI coding agent
  • Runs local language models on your machine
  • Native app keeps code and data local
  • Can crawl websites and export clean Markdown
  • Built by the creator of LLM Scout

Code Scout is a free native desktop AI coding agent that runs local or cloud models to automate repository tasks.

for
Developers who want AI-assisted coding while keeping code on their own machine.
pricing
free
license
open source
frankmedia/code-scout 0 0TypeScriptupdated 5 days ago
  • v0.1.36Code Scout v0.1.367 months ago
  • v0.1.34Code Scout v0.1.347 months ago
  • v0.1.30Code Scout v0.1.307 months ago
works withOllamaLM Studiollama.cppGitShell

Key features

6 features of Code Scout
  • Real tool access — Executes shell, file and git commands directly from the agent.
  • Repo-aware context — Stores project structure and notes in .codescout files for model reuse.
  • Planning-execution loop — Separates planning from execution to keep context small and focused.
  • Benchmark mode — Compares latency, output quality and cost across local and hosted models.
  • Visible heartbeat — Shows when the agent is running a tool, streaming output, or waiting on the model.
  • Native app — Built with Tauri 2 and Rust, runs on macOS, Windows and Linux without Electron.

Use cases

  • Automate repetitive refactoring across a codebase using local models.
  • Run tests, apply fixes, and commit changes without leaving the repository.
  • Benchmark different LLM backends to choose the cheapest or fastest for a task.

Code Scout vs alternatives

Code ScoutAugment CodeOpenCodeTabby MLbolt.diy
Best forLocal AI coding automationDeep context AI coding agentCLI-based AI codingEditor-embedded AI assistantFull-stack AI app generation
PricingFreeSubscriptionFreeFreeFree
DevHunt upvotes10000
LaunchedNov 2026————

Code Scout FAQ

Do I need to create an account to use Code Scout?+

No, the app is free to download and requires no sign-up.

Can Code Scout run entirely offline?+

Yes, you can point it at local models like Ollama, LM Studio or llama.cpp and keep all code on your machine.

What platforms are supported?+

Native builds are available for macOS (Apple Silicon and Intel), Windows and Linux.

How does the tool handle model costs?+

When using a cloud provider you supply your own API keys, so only the provider’s usage fees apply; local models incur no external cost.

Is the source code available?+

Yes, the source is public on GitHub.

What is the current development stage?+

Code Scout is in alpha as of October 2026, so expect rough edges and frequent updates.

Summarized by DevHunt from frankvitetta.net · Nov 17, 2026. Details may change; check the official site.