Free, open-source field service software for service businesses.
Requests, quotes, scheduling, crews in the field, invoicing and a client portal, in one app you own.

Website · Docs · Deploy on Railway · Self-host · What's inside · Develop · Full deployment guide

Crewtavo is for lawn care, paving, roofing, cleaning, HVAC and any trade that quotes work, schedules crews, works at a property and gets paid. It does what Jobber does, plus an SOP knowledge base, equipment maintenance, satellite map measuring, plan sets on jobs (lined up on the map to measure from), and connections to QuickBooks, Zoho, Claude and ChatGPT.

Every business runs its own instance: the app, its own database and its own files, in an account the business owns. Nobody else holds its data, and there's no per-seat subscription.

The website, crewtavo.com, has a 24-second tour, comparisons with Jobber and other apps, and the docs for business owners.

Deploy on Railway

The quickest way to run Crewtavo, and the one that suits a business without a developer.

1. Press Deploy on Railway above. Sign in to Railway, or create an account. Railway's free trial is enough to try Crewtavo.
2. Press Deploy. Railway creates your instance: the Crewtavo app, a Postgres database, and the services behind sign-in, data and file storage. Every password and key is generated for you, so there's nothing to fill in.
3. After a few minutes, open the address shown on the crewtavo service, create the owner account and set up your business. Then invite your team from Settings → People.

If something isn't ready yet, that address shows a Finish setting up page saying what to do, instead of an error.

Note

Cost. The business pays Railway directly. A new instance at rest uses about 0.36 GB of memory across its five services, roughly $3.60 a month of Railway usage. That fits Railway's free trial (30 days, $5 of credit) for trying Crewtavo out, and Railway's Hobby plan ($5 a month, including $5 of usage) to keep it running. Busy instances use more; check Railway's pricing for current numbers.

Your own domain: when you first sign in, Crewtavo asks to connect to Railway (you're still signed in there, so it's one press of Authorize), which is how it warns you before Railway's trial runs out. Right after you set up the business, it offers to set up your web address:

• Need a domain? Buy one on Railway. Railway sets it up, Crewtavo spots it and switches over, and there's nothing to add.
• Have one already (like app.yourbusiness.com)? Crewtavo adds it in Railway and shows the two DNS records to add where your domain is managed. On Cloudflare, it can add them for you. It checks the records as they go in and switches over by itself once the new address works.

Links already sent keep working, and the old address forwards to the new one. Skip it, and Settings → Web address does the same later, including the steps by hand.

Email: Railway blocks outgoing mail servers (SMTP) on its trial and Hobby plans, so Crewtavo sends email through Resend, which works on every plan: create a free Resend account and paste its API key under Settings → Email. On Railway's Pro plan, or when you self-host, you can instead connect the business's own mailbox there (Gmail, Google Workspace, Microsoft 365, Zoho, Yahoo, iCloud or any SMTP server). RESEND_API_KEY and EMAIL_FROM on the crewtavo service still work too, for any business that hasn't connected its own.

Backups: Crewtavo backs up its database every night, on every Railway plan, and keeps the last week's and month's. They're kept in the same Railway project, so download one from Settings → Your data now and then and keep it somewhere else. Photos and files aren't in a backup. Restoring one into a new instance is a few steps.

Important

Save JWT_SECRET (and ENCRYPTION_KEY, if you set one) and POSTGRES_PASSWORD somewhere outside Railway, such as a password manager. Without JWT_SECRET, a restored Crewtavo can't read its connected apps, the owner's key file stops working, and links already sent to clients stop opening; without POSTGRES_PASSWORD, the services can't sign in to a database volume you kept. In Railway they're under Variables: JWT_SECRET and ENCRYPTION_KEY on the crewtavo service, POSTGRES_PASSWORD on db.

Self-host

Crewtavo runs anywhere Docker does: your laptop, an office server or a VPS. It's the same five services the Railway template creates, described in docker-compose.yml.

You need Docker and Node.js 20 or newer (only to generate the secrets and start Docker; nothing to install).

git clone https://github.com/jviars/crewtavo.git
cd crewtavo
npm run instance

The first run writes .env.instance with fresh secrets, builds Crewtavo, starts everything and prints the address: http://localhost:8080. Open it and create the owner account.

Important

Keep .env.instance, and save a copy somewhere off this computer. It holds JWT_SECRET and POSTGRES_PASSWORD (and ENCRYPTION_KEY, if you set one). Without JWT_SECRET, a restored Crewtavo can't read its connected apps, the owner's key file stops working, and links already sent to clients stop opening; without POSTGRES_PASSWORD, the services can't sign in to the database volume.

npm run instance -- logs                           # follow what's happening
npm run instance -- stop                           # stop it; the data stays in Docker volumes
npm run instance                                   # start again, or update after a git pull
docker compose --env-file .env.instance down -v    # delete the instance and all its data

On a server with your own domain

Settings → Web address walks you through these steps with your domain filled in, and checks each one.

1. Point a DNS record for your domain (say ops.example.com) at the server.

2. In .env.instance, set APP_URL=https://ops.example.com, then run npm run instance again.

3. Put a web server with HTTPS in front of port 8080, on the same server. With Caddy, this whole Caddyfile gets a certificate and forwards everything (Caddy passes on each visitor's address in X-Forwarded-For, which Crewtavo's rate limits go by; see docs/deploy.md for other web servers):

   ops.example.com {
     reverse_proxy localhost:8080
   }
   

Only the crewtavo container is published, and only on the server itself (port CREWTAVO_PORT, default 8080, on CREWTAVO_BIND, default 127.0.0.1), so visitors come in through the web server. It forwards sign-in, data and files to the other containers over Docker's private network.

Backups: Crewtavo backs up its database every night; download one from Settings → Your data now and then and keep it on another computer (restoring one is npm run instance -- restore <file>). The data itself lives in two Docker volumes, crewtavo-instance_db-data (the database) and crewtavo-instance_storage-data (photos and documents). Back up both for the photos too, along with .env.instance.

On Railway, without the template

.railway/railway.ts describes the same instance as Railway Infrastructure as Code. In an empty Railway project:

cd .railway && npm install
railway link
railway config apply

Then set the two secrets it marks preserve() in the dashboard and generate a domain for the crewtavo service. deploy/README.md covers the template itself.

Optional features (texting, online payments, maps, AI, accounting) each take a key or two. Add them to the crewtavo service's variables on Railway, or to .env.instance when self-hosting. See Configuration.

What's inside

Schedule. Day and week time grids with a lane per crew, a month view, drag to move or resize, double-booking warnings, rain-day shifts and route order. The board keeps itself up to date as others make changes, and everyone can add their schedule to Google, Apple or Outlook Calendar. Calendars from Google, Zoho, Outlook, iCloud or any ICS link (or an uploaded calendar file) show alongside the visits, and Find a time steers around them.	Quotes. Line items from your price list, formulas and satellite map measurements, optional add-ons, deposits and promo codes. Clients approve and sign online.
Sales board. Every request and quote from first call to approval, with what's open, won and lost.	Finance. Collected, net and invoiced against the previous period, every payment, unpaid invoices by age, top clients and a yearly revenue goal, plus sales, work, team and client reports. Everything exports to CSV.
Equipment. Trucks, trailers and machines with hour meters and odometers, maintenance schedules by date, hours or miles, service history and costs.	Knowledge base. Your SOPs, with an editor, revision history and required reading for the crew.

In the field. My day on any phone: on my way, start and finish, gate codes, checklists with signatures, photos, materials used and a crew clock. It installs like an app on Android and iPhone, keeps working with no signal and catches up when the connection is back. For your clients. The Client Hub, with no account to create: approve quotes, pay invoices (with an optional tip), see the next visit, ask for changes, request work and message the office.

More screenshots

Measure. A paving takeoff drawn on the aerial.	Paving. The takeoff priced into a proposal.
Jobs. Every job and where it stands.	A job. Visits, line items and job costing.
A client. Properties, work, balance and contacts.	Invoices. What's owed, past due and collected.
An invoice. Payments and the printed copy.	Timesheets. The crew's hours, approved for pay.
Marketing. Campaigns, review asks and referrals.	Automations. Rules that follow up by themselves.
Apps. Built-in apps and connections.	Dark mode. Home after dark.

A visit on the crew's phone.

Home on the owner's phone.

And also:

• Jobs and invoicing: one-off and recurring jobs, multi-day work, change orders the client approves, batch invoicing, payment reminders, deposits, progress billing and retainage. Invoices go through Stripe for card and bank payments; invoices made in Stripe come in too, with the fees and bank deposits behind every payment. Refunds, write-offs for invoices that won't be paid, expenses and job costing, timesheets and pay.
• Plans and files: plan sets, specs, permits and addenda on requests, quotes, jobs and properties, up to 200 MB each. New versions replace old ones (with the history kept), the crew sees only current sheets on their phones, and clients can attach plans to a booking request. A site plan can go on the measuring map, lined up with the aerial by tapping two spots, to measure new paving from it.
• People and roles: General manager, Office manager, Operations manager, Foreman and Employee to start with, renamable, plus your own. A foreman runs their crew and never sees prices. A Subcontractor sign-in sees only their own jobs, their price on each and their paperwork. The database enforces every rule, not just the screens.
• Subcontractors: a record for each sub, whether or not they sign in, with their W-9, insurance certificates and licenses, expiry reminders, and upload links for subs without an account. Job offers they accept or decline in the field app (a lapsed certificate warns, and the offer records it), what's owed and paid, 1099-NEC totals by year, and a workers' comp audit report. Subs sign in with a link or a code by text or email, no password.
• Communication: email from your own mailbox, two-way texting through Twilio, a message center, and notifications in the app, by email, by text or as phone push, each person's choice.
• Automations and marketing: rules like "when a quote is sent, wait 3 days, then email the client", campaigns, promo codes, Google review requests and referrals.
• Claude and ChatGPT: a built-in connector that can look things up or run the business: import customers and recurring services (with a preview first, and no duplicates on a second run), manage the schedule, quotes, invoices and payments, read the money figures, invite people and change roles, approve time and mark people paid. Each person signs in with their own account; an assistant only reads until that person chooses what it can change, or gives it full access, with or without money (and it never does more than their role allows). Every call is logged.
• Pay: Timesheets → Pay sums each person's approved hours at their rate for a week or any dates, adds their piece pay (per square, per job or per day, from their rate card), flags overtime and unapproved time, downloads for your payroll service, and records who's been paid and how. Paid time is locked, and hours are paid once: where periods overlap, a later one owes only what the earlier didn't pay. Subs are paid by the job instead, under Subcontractors. Crewtavo doesn't move the money.
• Every trade: trade presets (packages/core/src/presets) load starter services, quote formulas, checklists, booking forms and SOP categories when a company signs up: plumbing, HVAC, electrical, cleaning, pool and spa, pest control, handyman, landscaping, lawn and exterior maintenance, paving and sitework, roofing and exteriors, or a general starting point for any other trade.

How it fits together

One instance per business: a Railway project, or docker compose on your own machine

Browser: React app (desktop and phone layouts, works offline in the field)
  └──► crewtavo  (the only public service: apps/api/src/server.ts)
         ├── the web app, and /api/* (Hono), acting as the signed-in user
         │     ├── QuickBooks Online, Zoho Books, Google Calendar (OAuth, tokens encrypted)
         │     ├── other calendars' ICS links (Zoho, Outlook, iCloud...), read every 15 minutes
         │     ├── Stripe, the company's mailbox or Resend, Twilio, Google Maps, Open-Meteo, Claude
         │     └── /mcp ◄── Claude and ChatGPT (OAuth through the instance's own sign-in)
         ├── background tasks: automations, reminders, recurring visits, weather holds
         ├── database migrations, applied on every start
         └── forwards to the instance's own services over the private network:
               /auth/v1     → auth      (Supabase Auth: sign-in, invitations, password resets)
               /rest/v1     → rest      (PostgREST, with row-level security)
               /storage/v1  → storage   (photos, receipts and documents)
                                all backed by db (Postgres)

The crewtavo service derives every service key from one shared secret and sets up the database itself, so a new instance needs nothing copied by hand.

Configuration

Everything an instance needs is generated. These optional settings turn on extra features. .env.example lists them all.

Optional settings

Name	Turns on
APP_URL	Your own domain. Settings → Web address sets it on Railway. Defaults to the instance's address.
ENCRYPTION_KEY	A separate key for app tokens and client links (32+ characters). Keep it stable once set: changing it disconnects every app and invalidates links already sent.
ANTHROPIC_API_KEY, AI_MODEL	AI rewriting and answers from your SOPs. Companies can also add their own key on the Apps page.
GOOGLE_MAPS_BROWSER_KEY, GOOGLE_MAPS_SERVER_KEY	Optional: Google satellite on the measuring map (Map Tiles API), and roof pitch suggestions (Solar API). Measuring needs no key: it uses free US and state imagery.
STRIPE_SECRET_KEY, STRIPE_WEBHOOK_SECRET	A Stripe account shared by every company on the server. Usually not needed: each business connects its own Stripe on the Apps page.
TWILIO_ACCOUNT_SID, TWILIO_AUTH_TOKEN, TWILIO_FROM_NUMBER	Text messages, and two-way texting.
RESEND_API_KEY, EMAIL_FROM	Email through Resend. Needed on Railway's trial and Hobby plans, which block SMTP; elsewhere, a default sender until the business connects its own mailbox.
QUICKBOOKS_CLIENT_ID, QUICKBOOKS_CLIENT_SECRET	An Intuit app for QuickBooks Online sync. Usually not needed: each business pastes its own Intuit app's keys on the Apps page.
ZOHO_CLIENT_ID, ZOHO_CLIENT_SECRET, ZOHO_ACCOUNTS_URL	A Zoho app for Zoho Books sync. Usually not needed: each business pastes its own Zoho app's keys on the Apps page.
GOOGLE_CLIENT_ID, GOOGLE_CLIENT_SECRET	Google Calendar sign-in: its calendars on the schedule, and visits on a Google calendar within minutes. Publish the OAuth app (Testing ends sign-ins after 7 days). Calendars added by ICS link (Zoho, Outlook, iCloud, Google's secret address) or uploaded as a file need nothing.
GOOGLE_PICKER_API_KEY	Import from Google Drive into Crewtavo's Drive, with GOOGLE_CLIENT_ID: Google's file picker, which lets Crewtavo read only the files people pick (drive.file, which needs no security review). A browser key for the Picker API, restricted to your domain and https://docs.google.com/*; the Picker and Drive APIs on, and your address under the client's JavaScript origins.
TURNSTILE_SITE_KEY, TURNSTILE_SECRET_KEY	Bot protection on public booking and referral forms.
CREWTAVO_SENTRY_DSN	Error reports to Sentry or a self-hosted GlitchTip, with personal data removed.
CREWTAVO_SCHEDULER	Set to off on extra replicas so background tasks run in one place.

docs/deploy.md has the details for each one (webhook addresses, the events to send, how error reports are scrubbed).

Develop

You need Node 20+ and Docker. The quick loop runs the API and the web app on your machine against a local Supabase:

npm install
npx supabase start               # local Supabase; applies supabase/migrations
cp .dev.vars.example .dev.vars   # paste the URL and keys that `supabase start` printed
npm run dev:api                  # API on :8787, restarting on changes, with the background tasks
npm run dev                      # web app on :5173, proxying /api to :8787
npm run seed:demo                # optional: a demo paving company with a week of work

Open http://localhost:5173 and sign in as demo@crewtavo.test (password crewtavo-demo), or create your own account. Email goes to Mailpit at http://localhost:56324; set MAILPIT_URL=http://localhost:56324 in .dev.vars and the app's own emails and texts land there too. Supabase Studio is at http://localhost:56323.

npm run typecheck      # every package
npm test               # core, API and web unit tests
npm run test:db        # migrations and row-level security tests in Docker Postgres
npm run test:e2e       # a job from new client to paid invoice, as four different roles
npm run test:browser   # the core flows clicked through in Chromium, with accessibility checks
npm run db:types       # regenerate database types after changing migrations
npm run screenshots    # retake the images in this README

test:e2e and test:browser need the local stack running and npm run seed:demo loaded. CI (.github/workflows/ci.yml) runs all of them, and also starts a whole instance with docker compose to check that it comes up ready for its owner.

Repository layout

Path	What's there
apps/web	The React + TypeScript + Vite web app
apps/api	The Hono API. src/server.ts serves it with the web app, runs background tasks and migrations, and fronts the instance's services (src/lib/stack.ts)
packages/core	Shared logic: map geometry, quote formulas, recurrence, time zones, roles, trade presets, database types
supabase	Migrations (schema, row-level security, database functions) and database tests
deploy	The instance's Postgres image and the Railway template guide
e2e	Playwright browser tests
scripts	Server build, migrations, npm run instance, the dev API, seeding, screenshots
docs	The deployment guide, the implementation plan, brand files and these screenshots

Status

Crewtavo works end to end: a whole instance starts on its own with docker compose, applies every migration, and goes from owner setup to a working business. The Railway template is new, so report anything that goes wrong on a first deploy. QuickBooks and Zoho sync are written but not yet tested against live accounts.

License

Crewtavo is free and open source under the GNU Affero General Public License, version 3 (AGPL-3.0-only), with the additional terms in NOTICE: no warranty, no responsibility for your data or for what you do with Crewtavo, and no rights to the Crewtavo name or logo for modified versions. If you run a modified Crewtavo for other people, the AGPL asks you to offer them its source code.

Whoever sets up a business in Crewtavo agrees to the Crewtavo Terms during setup: the business runs Crewtavo on its own accounts and is responsible for its data and its customers.

Copyright (C) 2026 Joshua Viars. Commercial licenses are available on request.

Contributing

Crewtavo doesn't accept contributions: pull requests are closed without review. Bug reports are welcome as issues and questions in Discussions, with no promise of support. See CONTRIBUTING.md.