Agents everywhere, and how to keep track of them
Welcome to the first Stackpick letter. Every Thursday: the new tools worth knowing about if you build with AI, what changed in our rankings, and one honest pick. This week's theme was hard to miss: people are running so many coding agents that a whole category of tools now exists just to keep an eye on them.
New this week
- Parallax (Coding agents): shows every coding agent you're running in one place. Useful once you have more than two going at once.
- Harness by Autonomous (Coding agents): runs and manages your coding agents across machines from one command center. For people running agents on more than one computer.
- shim-cli (Coding agents): shows what your coding agent sends out and lets you mask sensitive data. Worth a look if you point agents at code with secrets in it.
- Ultracite v8 (Coding agents): a linter and formatter made for code written with AI agents. A cheap way to keep agent output consistent.
- Lovable Growth Tools (App builders): Lovable now helps you grow the app you built, in the same chat. If you already build on Lovable, this is the next step after launch.
- NitroStack (Hosting and deploy): build, test and deploy MCP servers in one place. For anyone giving their agents custom tools.
What moved in the rankings
Nothing yet, because this is where we start. Every use case has one standard test (for coding agents: add email and password login to a small starter app, with tests passing), and every tool gets the same scorecard: does it do the job well (50%), speed and effort (20%), price for value (20%), and how easy it is to start (10%). Until a tool's test is done, it's marked "Test pending" and stays in our current pick order. Results land here first.
The pick: Supabase
Who it's for: anyone vibe coding an app that needs users and data.
Supabase gives you a Postgres database, login and file storage in one place, and most AI app builders and coding agents already know how to use it. That last part matters more than it sounds: when your agent has seen a tool thousands of times, it makes fewer mistakes setting it up.
The catch: row-level security is what keeps one user from reading another user's data, and AI-generated setups often get it wrong. Ask your agent to turn it on and test it.
Price: free tier to start. Test pending.
One question
What are you building right now? Hit reply and tell us. We read every answer, and it decides what we test first.