EvolutionAPI multi-instance

EvolutionAPI supports multiple instances (numbers) on the same installation. Each instance is a paired WhatsApp number, with its own isolated session and webhooks.

How it works

Each instance is identified by an instanceName. The API routes look like this:

POST /message/sendText/{instanceName}
GET  /chat/whatsappNumbers/{instanceName}
GET  /instance/connectionState/{instanceName}

Each number’s session is persisted in Postgres or Redis (configurable via DATABASE_PROVIDER).

Create a new instance

curl -X POST https://sua-evolution.rollinhost.com.br/instance/create \
  -H "apikey: SUA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "instanceName": "vendas",
    "qrcode": true,
    "integration": "WHATSAPP-BAILEYS",
    "webhook": {
      "url": "https://n8n.exemplo.com/webhook/vendas-in",
      "events": ["MESSAGES_UPSERT"]
    }
  }'

The response includes the qrcode.base64 for pairing.

Practical limits

Resource	Per instance	5 instances	20 instances
RAM	~400 MB	~2 GB	~8 GB
CPU (idle)	~2%	~10%	~40%
CPU (at peak)	~30%	varies	requires 8+ vCPU
Database	~50 MB/month	~250 MB	~1 GB

Horizontal scaling

For >50 instances, replicate EvolutionAPI across multiple servers and use a router based on instanceName:

                ┌─ evo-1.rollinhost.com.br (instances A-F)
NGINX/Traefik ──┼─ evo-2.rollinhost.com.br (instances G-M)
                └─ evo-3.rollinhost.com.br (instances N-Z)

Postgres can be shared across nodes (watch the connection pool).

Bans and best practices

• Keep an organic ratio between sent and received messages (avoid one-way blasting)
• Warm up new numbers: start with 50 msgs/day and double it every week
• Use a random delay between messages (1-3s), do not send in bursts

Next steps

• Connect a WhatsApp number
• Webhooks