跳到主要内容

Deploy to the cloud ☁️

Outgrown your laptop? Time to put cMind on real infrastructure. Good news: it's designed to scale out with almost no operator ceremony — no ZooKeeper, no leader election, just replicas and a database.

The one thing to know up front: the stateless tier (Web + MCP) runs happily on any container platform, but node agents need privileged Docker (they build and run cTrader containers). That rules out serverless runtimes like Azure Container Apps and AWS Fargate for the agents — run those on Kubernetes, a VM, or EC2 and point them at your Web URL.

Pick your path:

  • 🟦 Azure — Container Apps + Postgres Flexible Server (Bicep).
  • 🟧 AWS — ECS Fargate + ALB + RDS (Terraform).
  • ⎈ Kubernetes — the Helm chart, works on AKS / EKS / anywhere.
  • 📈 Scaling — how it all scales and self-heals once it's up.

Stateless tier (Web + MCP) run on any container platform; Postgres = managed database. Node agents need privileged Docker (DinD) — serverless container runtimes (Azure Container Apps, AWS Fargate) block it. Run agents on Kubernetes (kubernetes.md) or VM/EC2, point at Web URL.

CloudStateless tierDatabaseGuide
AzureContainer Apps (Bicep)Postgres Flexible Servercloud-azure.md
AWSECS Fargate + ALB (Terraform)RDS Postgrescloud-aws.md

Common prerequisites, both:

  1. Build + push three images to registry cloud can pull (cmind-web, cmind-mcp, cmind-node-agent).
  2. Pick secrets: DB password, owner email/password, discovery join token (≥ 32 chars) shared by Web app + every node agent.
  3. Deploy IaC (below), then bring node agents up separately (K8s/VM) with NodeAgent__MainUrl = deployed Web URL, NodeAgent__JwtSecret = join token.

Discovery, logging, probes behave same as local/K8s setups — see ../operations/node-discovery.md and ../operations/logging.md.