Sequo writes the plan and the decisions into files you keep in your repo, then hands you one prompt per step. Every prompt names the files to read first, so you stop being the one who remembers.
How it works #
Why not just CLAUDE.md or AGENTS.md? #
A markdown file works. Until the project outgrows it.
One file is the right call early. You write down the stack and a few decisions, the agent reads it, and everything holds.
Then the project grows. The file gets long, the decisions pile up, and the agent starts skipping it — no error, no warning, it just stops arriving with the context. You find out when finished work comes back rewritten.
From there you have three options. Keep the documents current by hand, which is a second job. Re-explain the project at the start of every session, which is the thing you were avoiding. Or let something else own it.
Sequo writes the documents as you go and names them in every prompt, so the file being read is never a question. And you get the plan itself: the whole build in dependency order, one prompt at a time.
Inside a step #
Every step splits in two: what your agent runs, and what only you can do.
Read docs/architecture.md and docs/decisions.md first.
Implement magic-link sign-in with Supabase Auth. Guard every project route. Leave the upload flow alone.
After the last step #
Shipping v1 isn’t the end of the plan. Pick what comes next and a new phase is planned on top of everything already built, with the documents updated to include it.
Pricing #
Build one thing free. Pay when you keep building.