Hey there, Im Mason 👋

I've spent the last 7+ years building software across startups and large corporations. I love the craft of engineering — but somewhere along the way, the ratio of building to shipping got out of balance.

I had too many projects and not enough hours in the day. Even with all the great AI tools out there (and I was paying for most of them), they still required me to sit down, prompt them, refine the output, iterate — over and over. As a husband and father, that time just isn't there.

I didn't need another AI assistant that waits for me. I needed something that works while I'm away.

That's why I built Arbiter. The idea is simple: front-load your thinking into well-written issues, then let Arbiter pick them up and do the work autonomously, using all those tools youre already paying for, but automating the boring parts like calling Github API's. When it's done, you review and merge. If something needs refinement, a few commands is all it takes.

I built Arbiter because AI coding tools are incredible — but you're still doing all the orchestration.

Claude Code, Cursor, Codex — they write great code. But you're still managing branches, running tests, creating PRs, and babysitting the process. The AI codes. You do everything else.

Arbiter isn't another AI coding tool. It's the orchestration layer that handles everything that isn't writing code — Git workflows, test execution, PR creation, error recovery, and retries. Your AI focuses on what it's good at. Arbiter handles the other 70%.

How it works:

Configure which issues Arbiter should pick up — filter by labels, assignees, or prioritize however you want
Arbiter hands matching issues to your AI and manages the full cycle
You get a draft PR ready for review

Small bug fixes, boilerplate, refactors, feature scaffolding, in-depth features — if it's in an issue, Arbiter can run it.

What's under the hood:

Bring Your Own AI — Claude Code, Codex, Cursor Agent, Ollama, LM Studio, and OpenCode all work on day one. Cloud or local, your call.
Local-first — Arbiter is a local CLI with a web UI that runs entirely on your machine. Your code is only shared with services you already control: GitHub (where your repo lives) and your chosen AI provider. We never send your code to our servers.
Self-learning (optional) — Arbiter learns from every PR, building knowledge of your codebase patterns. You approve or deny each learning.
Run your way — Click "Run Now" or set an auto-run interval. Built-in change, turn, and time limits mean no runaway cycles.
PR iteration — Comment on a draft PR with changes, Arbiter implements them automatically.
Auto-continuation — If a run hits its limits before finishing, Arbiter picks up right where it left off on the next run. No lost progress.
AI code review (optional) — Spin up a second AI worker to review PRs with configurable strictness. It leaves feedback, approves or requests changes, and applies a human-needed label when it can't make the call.
Observability — Built-in logging for Postgres and Elasticsearch, plus configurable webhooks on nearly every event.
No auto-renewal — Pay what you think is fair ($6.00+ /mo). When your license expires, you decide if it earned a renewal.

And more, check out the Docs to see some of the other cool features!

Arbiter began as a way for me to take back control of my passion, and Im hoping some of you can use it to do so too. Happy to answer any questions you have here.

PS: To celebrate launch week, I'm offering 50% off all licenses with code 'LAUNCHWEEK26' and giving away one free lifetime license every day on X and Discord (2 total per day) — something I won't normally offer for purchase. Come say hi!

Cheers,

- Mason