About Mechanical Advantage

Mechanical Advantage is the company behind Cog. Cog is the current product: autonomous software development that takes an issue from description to merged PR. It is the first of a family of mechanical parts — each a named product, each doing one thing well. More will follow.

Why Cog Exists

Developers already use CLI harnesses — Claude Code, Codex CLI, Gemini CLI, OpenCode — and they have never been more productive. The harness writes the code. The developer is still at the keyboard babysitting the rest: running the tests, opening the PR, reading review comments, fixing the CI failure, resolving the merge conflict, shepherding it over the line.

That middle is mechanical work. It takes judgment, but it does not take human judgment. Cog closes the loop — issue, code, PR, review, merge — so the developer’s hands-on time collapses to the two moments that actually need a person: describing what to build, and approving the result.

Our Principles

  • Harness-agnostic. Cog orchestrates Claude Code, Codex CLI, Gemini CLI, and OpenCode. We are not in the business of writing another coding agent; we are in the business of getting more out of the ones you already use.
  • Runs on your own hardware. No SaaS dashboard, no cloud execution. Cog runs on your machine, against your GitHub, with your harness subscription. Your code never leaves your environment except through a PR you approve.
  • Flat pricing. $1,000/month, unlimited repos. Not metered on tokens, not metered on agent-hours. You should be able to run Cog hard without watching a bill climb.
  • Developer-owned code. The output is a PR on your repo, written against your conventions, reviewable in the tools you already use. We never hold your code hostage and there is nothing to migrate off of.

Our Team

Human Workforce

David Weiss

Founder

David founded Mechanical Advantage to give AI agents the safe, reliable tools they need to act in the real world. He brings experience in software architecture, developer tooling, and product design — with a focus on building systems where agent mistakes are reversible, not catastrophic.

AI Workforce

Tintin

AI Workforce

Mac mini (2024), Apple M4, 16 GB unified memory, 256 GB SSD. macOS 26.3.1.

Thompson

AI Workforce

GEEKOM A6, AMD Ryzen 7 6800H, 16 GB DDR5, 1 TB NVMe. Windows 11 Pro.

Thomson

AI Workforce

GEEKOM A6, AMD Ryzen 7 6800H, 16 GB DDR5, 1 TB NVMe. Ubuntu 26.04 LTS.