Idea to Deployment
This is the workflow I use today for my personal projects. It takes a software product from a rough idea to a live product using two Claude tools: a copilot built as a Claude Project does the thinking with me, and Claude Code does the building while I review and approve.
Read the map from top to bottom. Click any box, including the two shaded containers, to open its explanation, then use Previous and Next to walk through all 14 parts in order.
Files to download in:
Key
Copilot (Claude Project): the thinking
Claude Code: the building
Everything else: start, end, people, Claude chat
Solid: the next step
Two-headed: back and forth
Dashed: optional path
Terms used on this page
- Claude Project:
- a workspace in Claude with its own instructions and files that every chat in it can use.
- Claude Code:
- Anthropic's coding agent. It works in a folder on your computer, reading and writing code and running commands.
- Repo:
- short for repository. The project's folder, tracked with version history on GitHub.
CLAUDE.md:- a text file in the repo with instructions Claude Code reads at the start of every session there.
- Plan mode:
- a Claude Code mode where it proposes a plan without changing any files.
- Skill:
- a set of instructions saved on your computer that Claude Code runs when you type its name, such as
/launch-readiness.
Principles
- Start building early. The workflow supports you; it isn't a gate.
- Keep docs few and current. The four files in the repo are enough, and the repo copies are the source of truth.
- One milestone at a time, each leaving the product working.
- Decide, then log it in
DECISIONS.md, so settled questions stay settled. - Add process only when something slips. If a step feels like overhead, skip it.