A Build Kit to help you build your own version of a personal bookmark manager using AI coding agents like Claude Code, Codex, or Gemini.
What you'll build
A private, single-user web app for saving and revisiting links. Paste a URL, the app auto-fetches the page title, and an AI summary fills in a few seconds later so future-you remembers why the link was worth keeping. Bookmarks are organized by a flat per-user tag system with sidebar filtering and search across title, URL, notes, and the AI summary.
Any saved bookmark can be emailed to one person in one click — the Share dialog opens with a friendly subject and body already pre-filled, including the URL and the AI summary, and any reply routes straight back to your account email.
The kit ships with the product spec, the milestone breakdown, the design system, and ready-to-paste prompts so an AI coding agent can build it for you in clear, sequenced stages.
Who it's for
The product is built for an individual reader — a knowledge worker, builder, researcher, or curious reader who reads a lot online and wants their own private "second brain" for links. Single-user only. Not a team workspace, not a public profile, not a social feed.
What it solves
- "I bookmarked this and now I can't remember why." An AI summary fills in for every bookmark a few seconds after saving, so future-you has real context before clicking in.
- "My bookmarks are a flat dumping ground." A flat per-user tag system with sidebar counts makes the collection navigable instead of a black hole.
- "I want to share this with one specific person without leaving the app." Every bookmark has a Share button that opens a pre-filled email — URL, summary, and a friendly default body — ready to send.
Key user flows
- A user can paste a URL into a clean form, see the title auto-populate, add notes and tags, and save in under a minute.
- A user can watch the AI summary fill in on the next page refresh — a clean "Summarizing…" placeholder while it's pending, the actual summary once it's ready.
- A user can click a tag in the sidebar to filter the list, click another tag to narrow further, and search across the filtered set.
- A user can rename or delete tags from a Manage Tags page; renames apply everywhere the tag is used; deleted tags are removed from every bookmark, but the bookmarks stay.
- A user can click Share on any bookmark, edit the pre-filled recipient, subject, and body, and send a plain-text email — replies route back to the user's account email.
How you'll build it
- Foundation and bookmark CRUD — set up the app's foundation (scaffold, auth, app shell, profile/settings, background job system) and ship bookmark CRUD with auto-fetched titles.
- Tag, search, and rediscover — add per-user tags with inline autocomplete, sidebar filtering, search across multiple fields, and a Manage Tags page.
- AI summaries fill in for every bookmark — wire up a background job that fetches the page text, calls the AI provider, and surfaces a 2–4 sentence summary on every bookmark with a clean three-state UI (pending / ready / failed).
- Share any bookmark via email — add a Share button that opens a pre-filled email dialog and sends a plain-text message via the mailer, with delivery and validation failure states handled inline.
What's included
README.md— orientation and how to use the kit_build_plan/PRD.md— the product requirements, data model, scope guardrails, and per-milestone breakdown with "Done when" criteria_build_plan/tech-stack.md— customizeable tech stack for your codebase_build_plan/design-system.html— visual reference (colors, typography, components — open in a browser)_build_plan/milestones/<N>-<slug>/prompt.md— one folder per milestone with a lean, ready-to-paste prompt that works with any AI coding agent
