Analytics cookies only (Google Analytics, PostHog). No ad networks, no cross-site tracking. See our privacy policy.
Updated Sep 2026
verified on Ubuntu 26.04 · Sep 2026Self-host NocoDB, the spreadsheet-style Airtable alternative, on your own VPS with Docker Compose, PostgreSQL, Redis and HTTPS, plus the backups and upgrades that keep it safe.
NocoDB puts a spreadsheet-style interface on top of a relational database. You get grid, kanban, calendar, gallery and form views, a REST API for every table, and sharing and permissions. You get all of it without writing SQL. It can keep its tables in its own PostgreSQL database, or you can connect an existing PostgreSQL or MySQL database and work on those tables in place.
It is the usual self-hosted answer to Airtable. The other one is Baserow, and NocoDB vs Baserow covers the differences. The short version: NocoDB is the better fit when you want a UI over a database you already run, and Baserow is the more self-contained product. Check the licence before you commit. The catalog lists NocoDB under the Sustainable Use License, not an OSI licence, so read it if you plan to resell or embed NocoDB.
On our test box (GCP e2-standard-2, Ubuntu 26.04, Docker, November 2026), a NocoDB stack idled at about 1,902 MB of RAM and used about 2.3 GB of disk for images and data. That was the stack with NocoDB, a worker, PostgreSQL and Redis. The catalog's minimum is 2 GB RAM. Treat that as the floor:
This guide assumes Docker Engine and the Compose plugin are installed, along with a firewall. If not, work through Docker & Compose on Ubuntu first.
Only SSH and the web ports need to be open. NocoDB itself will listen on loopback only:
sudo ufw allow OpenSSH
sudo ufw allow 80
sudo ufw allow 443
sudo ufw --force enable
sudo ufw status verbose
Paid link — we earn a commission if you shop through it.
NocoDB publishes a plain Compose stack (the quickstart-demo example in its
repository). It runs the NocoDB app, a background worker for imports,
exports and automations, PostgreSQL and Redis. The version below
follows that file, with three changes for a server:
127.0.0.1 so that only the reverse proxy can reach it.NC_SITE_URL to your public URL.Create the project directory and an .env file with generated values:
mkdir -p ~/nocodb && cd ~/nocodb
if [ ! -f .env ]; then
cat > .env <<EOF
POSTGRES_PASSWORD=$(openssl rand -hex 24)
NC_AUTH_JWT_SECRET=$(openssl rand -hex 32)
NC_SITE_URL=https://nocodb.example.com
EOF
fi
chmod 600 .env
The if guard stops you overwriting the database password on a second run.
Change the password after PostgreSQL has initialised and NocoDB can no longer
log in. Replace nocodb.example.com with your real hostname. NocoDB builds
invitation and password-reset links from NC_SITE_URL, so a wrong value
produces broken links.
Now the Compose file:
cd ~/nocodb
cat > docker-compose.yml <<'YAML'
services:
nocodb:
image: nocodb/nocodb:latest
restart: unless-stopped
environment:
NC_DB: "pg://db:5432?u=nocodb&p=${POSTGRES_PASSWORD}&d=nocodb"
NC_REDIS_URL: "redis://redis:6379"
NC_SITE_URL: ${NC_SITE_URL}
NC_AUTH_JWT_SECRET: ${NC_AUTH_JWT_SECRET}
NC_DISABLE_MUX: "true"
depends_on:
db: { condition: service_healthy }
redis: { condition: service_healthy }
volumes:
- nocodb_data:/usr/app/data
ports:
# Loopback only: Caddy is the public entry point.
- "127.0.0.1:8080:8080"
healthcheck:
test: ["CMD-SHELL", "wget -q --tries=1 --spider http://localhost:8080/api/v1/health || exit 1"]
interval: 30s
timeout: 5s
retries: 5
start_period: 30s
worker:
image: nocodb/nocodb:latest
restart: unless-stopped
environment:
NC_DB: "pg://db:5432?u=nocodb&p=${POSTGRES_PASSWORD}&d=nocodb"
NC_REDIS_URL: "redis://redis:6379"
NC_SITE_URL: ${NC_SITE_URL}
NC_AUTH_JWT_SECRET: ${NC_AUTH_JWT_SECRET}
NC_WORKER_CONTAINER: "true"
depends_on:
nocodb: { condition: service_healthy }
volumes:
- nocodb_data:/usr/app/data
db:
image: postgres:17.10
restart: unless-stopped
environment:
POSTGRES_USER: nocodb
POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
POSTGRES_DB: nocodb
volumes:
- postgres_data:/var/lib/postgresql/data
healthcheck:
test: ["CMD-SHELL", "pg_isready -U nocodb -d nocodb"]
interval: 10s
timeout: 5s
retries: 5
redis:
image: redis:7
restart: unless-stopped
volumes:
- redis_data:/data
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 10s
timeout: 5s
retries: 5
volumes:
nocodb_data:
postgres_data:
redis_data:
YAML
Bring it up. --wait blocks until the health checks pass, which takes a minute
or two on the first run while the images download:
cd ~/nocodb
docker compose up -d --wait
docker compose ps
Check that the app answers on loopback. Expect a 200:
curl -s -o /dev/null -w '%{http_code}\n' http://127.0.0.1:8080/api/v1/health
If you'd rather not maintain a Compose file, NocoDB also ships a single-server
installer, noco.sh, that generates this stack plus Traefik and Let's Encrypt
for you. It is the method shown on the NocoDB page. The
hand-written file above is easier to read, back up and diff.
Point an A record for nocodb.example.com at the server's public IP. Then
put Caddy in front of 127.0.0.1:8080. The full setup is in
Automatic HTTPS with Caddy. The site
block is:
nocodb.example.com {
reverse_proxy 127.0.0.1:8080
}
Caddy passes WebSocket upgrades through without extra configuration. Reload
Caddy, then open https://nocodb.example.com.
The first account to sign up becomes the super admin. Do it straight after the site goes live, before anyone else finds the URL. Then:
NC_INVITE_ONLY_SIGNUP. After
that, new users only get in through an invitation.8080 to loopback. Do
not change it to 8080:8080. Docker-published ports bypass ufw, so that
change would expose NocoDB over plain HTTP..env. It holds the database password and the JWT secret. It is
chmod 600 and should never go into git.NC_DISABLE_TELE: "true" to both app services to turn it off.Your tables and NocoDB's metadata live in PostgreSQL. Uploaded attachments
live in the nocodb_data volume. Back up both.
A logical dump of the database is safe to take while the stack is running:
cd ~/nocodb
mkdir -p backups
docker compose exec -T db pg_dump -U nocodb -d nocodb | gzip > backups/nocodb-db-$(date +%F).sql.gz
ls -lh backups/
Attachments come from the data volume. Compose prefixes the volume name with
the project directory, so check the real name with docker volume ls:
cd ~/nocodb
docker run --rm -v nocodb_nocodb_data:/data:ro -v "$PWD/backups":/backup alpine \
tar czf /backup/nocodb-data-$(date +%F).tar.gz -C /data .
Copy backups/ and your .env off the server. Without the .env file
you have the data but not the secrets that match it. To restore, start a fresh
stack with the same .env. Load the dump with
gunzip -c … | docker compose exec -T db psql -U nocodb -d nocodb, then
unpack the attachments archive into the new volume.
cd ~/nocodb
docker compose pull
docker compose up -d --wait
Take the pg_dump backup above first, every time, because a new version can
change the database schema when it starts. latest follows current releases.
If you'd rather choose when to upgrade, pin nocodb/nocodb to a version tag
from its releases page and change the tag deliberately. Read the release notes
before a big version jump.
docker compose up --wait times out on nocodb. Read the logs with
docker compose logs nocodb --tail 100. The usual cause is a database
connection error, because the password in NC_DB no longer matches the one
PostgreSQL was first initialised with. PostgreSQL only reads
POSTGRES_PASSWORD when its volume is empty, so changing .env later does not
change the password.
Invitation or reset links point to localhost. NC_SITE_URL is unset or
wrong. Fix it in .env and run docker compose up -d.
Imports hang. Imports and exports run on the worker container. Check that
it is running (docker compose ps) and read docker compose logs worker.
The box runs out of memory. The stack idles around 1.5 GB. Large CSV imports push it higher. Move to 4 GB, or add swap as a stopgap.
You're done when:
https://nocodb.example.com loads with a valid certificate;docker compose restart;pg_dump sits somewhere other than this server.From there, connect an existing database as a data source, or try the REST API with a scoped API token. If you're still choosing, compare it with Baserow on a VPS or look at Directus for a headless data platform. For ranked hosts, see Best VPS for databases.
We estimate about 30 minutes of hands-on time for "How to Deploy NocoDB on a VPS"; this is an editorial estimate, not a measurement.
The "How to Deploy NocoDB on a VPS" guide was verified end-to-end on Ubuntu 26.04 (Sep 2026).