crm

Sales CRM

A CRM and sales pipeline for founders who do their own selling: leads, companies, deal stages, follow-ups, meeting history, and sales forecasts. Built with vanilla JavaScript on Puter, so all data lives in the signed-in user's own Puter cloud. No app server, no database to run.

Features

• Saved views. The filter combinations you work from every day - "cold leads in SaaS", "deals closing this quarter, biggest first" - saved by name on Leads, Companies, and the deals list, as a row of chips above the filters. One click applies one; one can be the view the page opens with; rename, reorder, and delete are in the chip's own menu. A view is a filter set, not a snapshot, so it re-runs every time - and a condition that no longer exists (a deleted tag, a removed stage) is ignored and said out loud rather than quietly emptying the list.

• Command palette. Cmd/Ctrl+K (or /) searches every contact, company, deal, follow-up, logged activity, and saved view at once, ranked and grouped, with quick actions and full keyboard navigation.

• Pipeline board. Deals move through New, Qualified, Proposal, Negotiation, Won, and Lost. Drag cards between stages on desktop, or use the per-card move menu (works with keyboard and touch, and doubles as the accessible path).

• Board or list. The same deals as a Kanban board or a sortable table (deal, company, stage, value, expected close, last updated), with shared filters: text search, stage, and close-date window. On the board, Won and Lost collapse to a total while you work the open stages.

• Your funnel, your stages. Rename, reorder, add, and remove open stages in Settings (up to eight); Won and Lost stay pinned as the terminals. Removing a stage that holds deals asks where they should go.

• Stage history. Every stage change is recorded, so a deal page shows "In Proposal for 12 days" and the journey that got it there. The board marks deals that have sat too long (threshold in Settings), the dashboard lists the most stalled, and the forecast reports average days per stage.

• Leads and companies. People and accounts with detail pages that gather every deal, follow-up, and logged interaction in one place.

• Follow-ups. Dated reminders tied to a lead or deal, with an optional time, grouped into overdue, today, and upcoming. The nav badge shows what needs eyes now.

• Playbooks. Save the cadence you run every time - recap the same day, nudge in three days, check in after two weeks, each step optionally carrying an email template - and apply it to a lead or deal in one click. A preview shows the dates before it commits, the whole application is one Undo, and a playbook can offer itself when a deal reaches a stage (it always asks, and only when nothing else is booked).

• A week you can see. Follow-ups as a month grid or a week strip beside the list: each day holds its items with a count, tinted for overdue and today, with open deal close dates riding along as the other dated commitment. Drag a follow-up to another day to reschedule it, or tap it (the keyboard and touch path) - one Undo either way. The list stays the default, and a phone gets the week as a stack of days rather than a shrunk month.

• Reminders that leave the app. Opt in to browser notifications for a daily digest and timed nudges while a tab is open, and export open follow-ups (plus deal close dates) as an .ics file for the calendar you already check.

• Going cold. Last-contact dates are derived from logged activity and shown on lead rows, lead pages, and deal pages. A Going cold filter on Leads ranks the quietest first, and the dashboard surfaces both drifting leads and open deals with no next step, each with one-click actions.

• Meeting history. Meetings, calls, emails, and notes logged against contacts, companies, and deals.

• Activity log. Every touch in one place, newest first, with filters for type, date window, and the record it was logged against, plus search over what you wrote. Set a weekly touch target in Settings - overall or per type ("8 calls, 3 meetings") - and the log and the dashboard track this week against it, with pace for the day of the week.

• Win/loss reasons. Closing a deal asks why (skippable in one click), and the Forecast page turns those answers into a report: why deals close, win rate by lead source and industry, average won vs lost deal size.

• The whole buying committee. A deal links the two to four people who actually decide it, each with a role - champion, decision maker, influencer, blocker - and its page lists them with their last-contact dates. Its timeline gathers what was said to any of them, going-cold reads the freshest touch across all of them, and the dashboard chases open deals where nobody is named as the person who signs.

• Custom fields. Define your own fields per record type in Settings - text, number, date, choice, or yes/no - and they appear on the add and edit forms and on detail pages. Choice and yes/no fields join the filter rows on Leads, Companies, and the deals list, and every field exports as its own CSV column and imports back by name. Changing a field's type converts what is stored or refuses; it never drops a value.

• Subscriptions, priced properly. A deal bills one-time, monthly, or annually over a contract term, and MRR, ARR, and total contract value are derived from that - never stored twice. The pipeline values a $2k/mo deal at its whole contract, the forecast spreads it across the months it actually bills instead of spiking the close month, and the monthly goal counts what billed in that month.

• Where deals leak. A conversion funnel on the Forecast page, derived from stage history: per stage, how many deals ever reached it, how many advanced, how many died there and what share of all losses that is, plus entry-to-won overall. Read it for all time, this quarter, last quarter, or the year. Deals that closed before stage tracking are excluded and counted, never passed off as instant conversions.

• Forecast. A month-by-month chart of weighted pipeline (deal value times stage probability) against best case, with won revenue alongside. Stage weights are tunable in Settings as real conversion rates emerge.

• Monthly revenue goal. Set the number you sell against in Settings and the dashboard turns won revenue into progress against it, with pace for the day of the month. The forecast draws the goal across the months ahead and reports pipeline coverage.

• Dashboard. Open pipeline, MRR, weighted forecast, won this month, win rate, what needs attention today, and recent activity.

• Light, dark, and system themes (system is the default). Responsive, app-like layout: sidebar on desktop, tab bar and floating action button on mobile.

• AI assist, on Puter's own models. Three buttons, each one pressed on purpose: paste a signature or some scribbles into a new lead or company and have the fields read out of it, draft a first pass of an email into the same editable body a template fills, and catch up on a long history. Every result lands in a field you can edit and nothing is saved on its own. One switch in Settings turns it all off, and it stays off wherever puter.ai isn't available - the record you press a button on is the only thing that ever leaves your own cloud.

• CSV in and out. Import leads, companies, or deals from a spreadsheet with auto-detected column mapping, a preview, and merge-by-email (or by name) so a re-import updates instead of duplicating. Deal billing and contract term ride along, and a file without those columns imports as one-time. Export any record type as CSV alongside the full JSON export.

• First-run onboarding with optional sample data, JSON export/import, and a guided empty state on every view.

• Works with no connection. A service worker caches the app shell, so an installed copy launches offline instead of showing nothing, and every collection keeps a durable local copy in IndexedDB - so a cold start with no network opens your own data. A write the cloud won't take is queued and replayed when it answers: the shell says "waiting to sync", Settings lists what is still on its way, and a save that landed locally says so rather than claiming more. If another device changed the same collection meanwhile, the queued write is held back and you choose which version stands - it is never decided quietly.

• Forgiving deletes. Every delete offers Undo and lands in Recently deleted (Settings) for 30 days, restoring its links on the way back. Import, sample-data load, and erase take a one-slot snapshot first, so a bad bulk operation is one click from being reversed.

Stack

• Vite for dev server and build
• Puter.js for auth, key-value storage, files, and (optionally) the built-in AI models
• Lucide for icons
• Vitest for tests
• No frontend framework

Development

npm install
npm run dev        # dev server
npm test           # run the test suite
npm run build      # production build to dist/

The real app (index.html) requires a Puter session. For local development without one, open /mock.html, which loads dev/puter-mock.js: an in-memory Puter SDK mock persisted to localStorage. /dev/phone.html wraps the mock in a 390px iframe for quick phone-size checks. Neither is part of the production build.

Storage

Everything lives in Puter KV, one key per collection (crm:settings:v1, crm:companies:v1, crm:contacts:v1, crm:deals:v1, crm:activities:v1, crm:tasks:v1), plus crm:trash:v1 for 30-day undo and crm:backup:v1 for the pre-import snapshot. A founder-scale CRM fits comfortably in KV, keeps reads to single keys, and makes export/import exact.

Each collection is also mirrored to IndexedDB (src/lib/outbox.js): that copy is what an offline cold start reads, and the queue beside it holds any write KV hasn't taken yet, with the value KV last had so a replay can tell "nothing changed" from "another device wrote".

License

This repository, including all its contents, sub-projects, modules, and components, is licensed under MIT unless explicitly stated otherwise. Third-party libraries included in this repository may be subject to their own licenses.