dFlowvsCoolify

COMPARE · COOLIFY

dFlow vs Coolify

Coolify is an open-source, self-hostable PaaS. It runs Docker workloads on servers you connect over SSH, and it now has an MCP server. dFlow runs on Kubernetes, from a CLI or a coding agent. Pick by what you want to operate.

Rival facts checked Nov 19, 2026. Sources at the end.

your app todayone command eachdFlow on Kubernetes
01

Who should pick which

Pick dFlow if

  • You want Kubernetes underneath, without running the control plane.
  • You want destructive agent tools to need a typed name, not only a token scope.
  • You want each release kept, with a CLI rollback by release ID.

Pick Coolify if

  • You want to self-host the whole platform, open source.
  • You run Docker or Docker Compose on your own VPS or bare metal.
  • You want 1-click services from Compose templates.
02

Side by side

FeaturedFlow beta, Nov 2026Coolify
Where apps runKubernetes. Our shared cluster, or yoursDocker on servers you connect over SSH 1
Deploy fromA GitHub repo (apps:build --repo) or an image (apps:deploy --image)A Git repo, a Dockerfile, a Docker Compose file or a prebuilt image 1
CLIThe dflow CLIThe coolify CLI for the Coolify API 2
Agent accessHosted MCP at k8s.dflow.sh/mcp, plus 4 skillsAn MCP server at /mcp over Streamable HTTP, with a Bearer token. Off by default on self-hosted 3
Guardrails for agent actionsDestructive tools need a typed argument, such as confirm_name or confirm_hostToken scopes. delete is separate from write, and MCP hides tools the token can't use 3
Rollbackdflow apps:rollback APP --release IDRollback is a deploy operation; the token needs deploy 3
PostgresPostgres 16, 17 or 18. On-demand backups and restoreSupported databases, with persistent storage and backup configuration 1
Cron and jobsdflow jobs:scheduleScheduled tasks with cron expressions 4
Your own servers or clusterRegister your own Kubernetes cluster from the CLIYes. Any server you connect over SSH 1
Self-host the platformNo. dFlow stays hosted. dflow-server cluster:create installs k3s or kubeadm on a machine an operator already has. It does not publish a self-hosted control plane.Yes. Open source, or Coolify Cloud as a managed control plane 1
StatusInvite-only beta. No SLA. Pricing not setGenerally available
03

What is the same

Both have a CLI and an MCP server, deploy from a repo or an image, run Postgres, and let you use your own machines.

04

Move from Coolify to dFlow

  1. Request beta access and install the CLI.
  2. Deploy the image you already build with dflow apps:deploy APP --image IMAGE.
  3. Create a Postgres database and link it to the app.
  4. Copy your data with standard Postgres tools. Get the connection details with dflow db:credentials.
  5. Set your env vars and secrets.
  6. Set the route, check the logs, then point your DNS to dFlow.
05

Sources

  1. 1Coolify Docs: What is Coolifycoolify.io/docs/get-started/introduction
  2. 2GitHub: coollabsio/coolify-cligithub.com/coollabsio/coolify-cli
  3. 3Coolify Docs: Set up MCPcoolify.io/docs/mcp/setup
  4. 4Coolify Docs: Cron syntaxcoolify.io/docs/knowledge-base/cron-syntax

Logo: Simple Icons (CC0). All trademarks belong to their owners.

06

More comparisons

Get a beta invite.

The beta is invite only. Tell us what you deploy and which agent you use. We reply by email.

Join the beta

Open to the first 100 users.

Book a design-partner call

Invite only. Read the beta terms.