omg.dev review: run AI coding agents on your own machine
omg.dev (formerly lfg) runs AI coding agents on your own machine. Local control vs managed agents, and when the install overhead is worth it.
What it actually does
omg.dev is a GitHub-hosted project for running AI coding agents on your own machine. It was previously published as BennyKok/lfg and has been renamed. The core idea is local execution: agents that assist with coding work stay on hardware you control rather than depending entirely on a remote hosted service.
In practical terms, that means you install and operate the stack yourself. The project sits in the growing set of tools aimed at developers who want agent-style help (planning steps, editing code, running commands in a loop) without sending every prompt and file to a third-party cloud workspace by default. Exact agent behaviours, supported models, and integration points are defined in the repository; this review does not invent a feature list beyond what the public description states.
Because it is machine-local, setup, model choice, and ongoing operation fall to you. That can include pulling models, configuring API keys where a remote model is still used, and managing the runtime environment. The value proposition is control and proximity to your codebase, not a polished multi-tenant product with a marketing site full of screenshots.
Who it is for
This is aimed at developers and technical users who already work comfortably with local tooling, terminals, and GitHub projects. If you prefer agents that can see and act on a real project directory on your laptop or workstation, and you are willing to install and maintain that stack, omg.dev is in the right category.
It suits people who care about keeping code and context on their own machines where possible, or who want to experiment with coding agents without committing to a fully hosted IDE-style agent product. Indie hackers, open-source contributors, and engineers evaluating local agent workflows are the natural audience. Familiarity with reading a README, resolving dependency issues, and updating a repo when the project moves quickly will help.
Teams that already standardise on self-hosted or air-gapped-friendly developer tools may also look at it as one option among several local agent runners, provided someone on the team can own installation and upgrades.
Who should skip it
Skip omg.dev if you want a browser-only product with accounts, billing, and support baked in. Non-developers, or developers who only want a managed chat-and-edit experience inside a hosted IDE, will find the “run it on your own machine” model a poor fit.
If you need formal vendor security reviews, SLAs, compliance paperwork, or a procurement-friendly contract, a personal GitHub project is the wrong place to start. The same applies if you cannot or will not install software locally, or if your organisation blocks unsigned or self-hosted agent tooling.
Anyone expecting a finished consumer app with one-click installers for every platform and no terminal work should look elsewhere. People who only need occasional autocomplete or simple chat about code may find lighter editor extensions enough without standing up a full local agent environment.
Pricing and catch
Pricing is unknown from the source listing. The project is published publicly on GitHub under BennyKok/omg.dev; many such tools are free to clone and run, with costs limited to your own hardware, electricity, and any model API usage you configure. That is not a confirmed price list. Check the repository for licence terms, optional paid services, or model requirements before you rely on it.
The practical catch is operational: you own setup, updates, model access, and troubleshooting. Local agents can still call external model APIs depending on how you configure them, so “on your own machine” does not automatically mean fully offline or zero data leaving your network. Read the project docs for what is sent where. Stability and documentation quality can vary on fast-moving personal projects; star counts and recent commits are signals to inspect, not guarantees.
There is no claim here of specific certifications, enterprise tiers, or guaranteed uptime. If those matter, treat them as unknown unless the maintainer states them clearly in the repo.
Bottom line
omg.dev is a straightforward proposition: run AI coding agents on hardware you control, via an open GitHub project that previously lived under the lfg name. It is useful when you want local agency over coding agents and accept the maintenance that comes with self-hosted developer tools. It is not a managed marketplace app with polished onboarding, fixed pricing pages, or enterprise assurances.
Evaluate it by cloning the repository, reading the current README, and testing on a non-critical project. Compare it with other local agent runners and with hosted alternatives according to how much control, privacy, and ops work you actually want. If the install path is clear and the agent loop fits your workflow, it can earn a place in a personal toolkit. If you need something that “just works” behind a login button, choose a different category of product.
For buyers browsing a marketplace of AI apps, the honest filter is simple: technical comfort with local installs, yes; demand for turnkey SaaS, no. Pricing and full capability detail remain whatever the maintainer documents on GitHub at the time you check—unknown until verified there.
Related reading
Erhan Timur, Founder, Digital by Default
Enjoyed this article?
Subscribe to our Weekly AI Digest for more insights, trending tools, and expert picks delivered to your inbox.