diff --git a/README.md b/README.md
new file mode 100644
index 0000000..68445a0
--- /dev/null
+++ b/README.md
@@ -0,0 +1,243 @@
+# vtstats server
+
+## Deploy
+
+We use podman to simplify the deployment.
+
+First, put these files on your linux machine:
+
+<details>
+<summary>
+<code>/etc/containers/systemd/vtstats-database.container</code>
+</summary>
+
+```ini
+[Unit]
+Description=vtstats postgresql database
+
+[Container]
+Image=docker.io/groonga/pgroonga:latest-debian-15
+Exec=-c shared_preload_libraries='pg_stat_statements' -c pg_stat_statements.max=10000 -c pg_stat_statements.track=all
+ContainerName=vtstats-database
+IP=10.88.0.10
+AutoUpdate=local
+Network=podman
+LogDriver=journald
+EnvironmentFile=/path/to/your/vtstats.env
+Volume=/path/to/your/postgres:/var/lib/postgresql/data
+
+[Service]
+Restart=always
+TimeoutStartSec=900
+
+[Install]
+WantedBy=multi-user.target default.target
+```
+
+</details>
+
+<details>
+<summary>
+<code>/etc/containers/systemd/vtstats-api.container</code>
+</summary>
+
+```ini
+[Container]
+Image=ghcr.io/vtstats/server:latest
+Exec=api
+ContainerName=vtstats-api
+IP=10.88.0.11
+AutoUpdate=registry
+Network=podman
+LogDriver=journald
+EnvironmentFile=/path/to/vtstats.env
+
+[Service]
+Restart=always
+TimeoutStartSec=900
+
+[Install]
+WantedBy=multi-user.target default.target
+```
+
+</details>
+
+<details>
+<summary>
+<code>/etc/containers/systemd/vtstats-worker.container</code>
+</summary>
+
+```ini
+[Container]
+Image=ghcr.io/vtstats/server:latest
+Exec=worker
+ContainerName=vtstats-worker
+IP=10.88.0.12
+AutoUpdate=registry
+Network=podman
+LogDriver=journald
+EnvironmentFile=/path/to/vtstats.env
+
+[Service]
+Restart=always
+TimeoutStartSec=900
+
+[Install]
+WantedBy=multi-user.target default.target
+```
+
+</details>
+
+<details>
+<summary>
+<code>/path/to/vtstats.env</code>
+</summary>
+
+```env
+# server
+SERVER_HOSTNAME=
+SERVER_ADDRESS=
+
+# database
+POSTGRES_PASSWORD=
+POSTGRES_USER=
+POSTGRES_DB=
+DATABASE_URL=
+
+# youtube
+YOUTUBE_PUBSUB_SECRET=
+YOUTUBE_API_KEYS=
+INNERTUBE_API_KEY=
+INNERTUBE_CLIENT_NAME=
+INNERTUBE_CLIENT_VERSION=
+
+# bilibili
+BILIBILI_COOKIE=
+
+# telegram
+TELEGRAM_BOT_TOKEN=
+TELEGRAM_SECRET_TOKEN=
+
+# discord
+DISCORD_APPLICATION_ID=
+DISCORD_APPLICATION_PUBLIC_KEY=
+DISCORD_BOT_TOKEN=
+
+# s3
+S3_PUBLIC_URL=
+S3_HOST=
+S3_REGION=
+S3_BUCKET=
+S3_KEY_ID=
+S3_ACCESS_KEY=
+
+# admin api
+ADMIN_USER_EMAIL=
+GOOGLE_CLIENT_ID=
+
+# twitch
+TWITCH_CLIENT_ID=
+TWITCH_CLIENT_SECRET=
+TWITCH_WEBHOOK_SECRET=
+```
+
+</details>
+
+Then start the database and run migration
+
+```shell
+sudo systemctl daemon-reload
+sudo systemctl start vtstats-database
+sudo podman run -it -e DATABASE_URL=YOUR_DATABASE_URL ghcr.io/vtstats/server:latest database-migrate
+```
+
+Finally, start the api server and job worker:
+
+```shell
+sudo systemctl start vtstats-api
+sudo systemctl start vtstats-worker
+
+# check status
+sudo systemctl status 'vtstats-*'
+
+# view log
+sudo journalctl -u vtstats-api --since '1 min ago' -f
+```
+
+## Update
+
+```shell
+sudo podman pull ghcr.io/vtstats/server:latest
+sudo podman run -it -e DATABASE_URL=YOUR_DATABASE_URL ghcr.io/vtstats/server:latest database-migrate
+sudo podman auto-update
+```
+
+## Grafana
+
+<details>
+<summary>
+<code>/etc/grafana-agent.yaml</code>
+</summary>
+
+```yaml
+server:
+  log_level: info
+
+metrics:
+  wal_directory: /tmp/agent
+  global:
+    scrape_interval: 60s
+    remote_write:
+      - url: your-url
+  configs:
+    - name: web
+      scrape_configs:
+        - job_name: web
+          static_configs:
+            - targets: ["10.88.0.11:9000"]
+    - name: worker
+      scrape_configs:
+        - job_name: worker
+          static_configs:
+            - targets: ["10.88.0.12:9000"]
+
+logs:
+  configs:
+    - name: journal
+      clients:
+        - url: your-url
+      positions:
+        filename: /tmp/agent/positions.yaml
+      scrape_configs:
+        - job_name: journal
+          journal:
+            json: false
+            max_age: 6h
+            path: /var/log/journal
+          relabel_configs:
+            - action: keep
+              source_labels: ["__journal__systemd_unit"]
+              regex: .*vtstats.*
+            - source_labels: ["__journal__systemd_unit"]
+              target_label: unit
+          pipeline_stages:
+            - json:
+                drop_malformed: true
+                expressions:
+                  level: level
+                  timestamp: timestamp
+                  message: message
+            - timestamp:
+                source: timestamp
+                format: RFC3339Nano
+            - labels:
+                level:
+            - output:
+                source: message
+
+integrations:
+  node_exporter:
+    enabled: true
+```
+
+</details>