Status: BETA · INVITE ONLY

Ship without a platform team.

repoagentk8s · managedbuilddeploy
FIG. 00 · repo → agent → cluster

Deploy to Kubernetes from the dflow CLI or your coding agent. We do the cluster work. You can roll back any release.

Open to the first 100 users.

$curl -fsSL https://github.com/dflow-sh/dflow/releases/latest/download/install.sh | sh

Invite only. We verify your email before your first deploy.

01The problem

Your team became a platform team.

You wanted to deploy an app. Now you own YAML, ingress, certs, CI and a pager. The cluster became the product. We think that work is our job, not yours.

  • You spent Friday night in CloudWatch.
  • The pipeline passed. Staging went down.
  • A green pipeline is not a good release.
YAMLingresscertsCIpager

02How it works

From zero to first deploy.

  1. Step 1

    Install the CLI

    One script installs the CLI. dflow upgrade gets each new version.

    $ curl -fsSL https://github.com/dflow-sh/dflow/releases/latest/download/install.sh | sh
  2. Step 2

    Deploy your app

    Point dFlow at a container image. dFlow deploys it to the cluster.

    $ dflow apps:deploy api --image ghcr.io/you/api:1.0
  3. Step 3

    Route and logs

    Get a hostname with automatic TLS. Then read the pod logs.

    $ dflow apps:route:set api$ dflow apps:logs api
  4. Step 4

    Roll back

    dFlow keeps your releases. Roll back to an earlier one.

    $ dflow apps:releases:list api$ dflow apps:rollback api --release rel_…

Every step is in the docs. Read the quickstart →

01install02deploy03route + logs04rollbackFIG. 01 DeployFIG. 02 Rollback

03Agents deploy too

Your coding agent can deploy too.

Connect your agent through hosted MCP or skills. Then ask it to deploy. For a delete or a restore, it must name the target.

A deploy fails at 3 AM. Your agent reads the errors. Should it roll back on its own? First, it must know how to undo the change.

Destructive actions require explicit confirmation.

$ npx skills add dflow-sh/dflow https://k8s.dflow.sh/mcp (Authorization: Bearer $DFLOW_TOKEN)
agentCLIMCPskillsdeleteconfirm_name

04EXAMPLE SESSION

The whole thing is commands.

$curl -fsSL https://github.com/dflow-sh/dflow/releases/latest/download/install.sh | sh
$dflow apps:deploy api --image ghcr.io/you/api:1.0
$dflow apps:route:set api
$dflow apps:logs api
$dflow apps:releases:list api
$dflow apps:rollback api --release rel_…

Six commands take an app from an image to a rollback.

05What's in the box

Included in the beta.

Managed Postgres

Postgres 16, 17 or 18. Back it up on demand, and restore.

db:postgres:create · db:backups:create

Volumes

Storage for apps that must keep their files.

volumes:create

Jobs and cron

Create jobs or schedule them. Read their logs.

jobs:schedule

Routes and TLS

A platform hostname, up to 10 custom domains and automatic TLS.

apps:route:set

Write-only secrets

You set a secret. Nobody can read it back.

Templates

Install umami or hasura. More templates are in development.

templates:install

06Bring your own cluster

Outgrow ours? Attach your own cluster.

Start on our shared cluster. Need more control? Register your own Kubernetes cluster. Your deploy commands stay the same. We store the kubeconfig encrypted.

$ dflow clusters:register --name NAME --region REGION --kubeconfig PATH

Cluster registration works from the CLI only, not MCP.

Own cluster →

appsharedyoursdflowkubeconfig · encrypted

07Trust

How we handle your secrets.

  1. Secrets are write-only. We encrypt them with AES-256-GCM and inject them as a Kubernetes Secret.
  2. Plain env vars are not secrets. We don't encrypt them. Put keys in secrets.
  3. dFlow redacts bearer tokens, URL passwords and kubeconfig secrets in logs.
youencrypteddatabasek8s SecretpodGET /v1/apps Authorization: BearereyJhbGciOiJIUzI1NiIsInR5

07·1Compliance

Status as stated on the dFlow trust center, which covers dFlow Cloud (app.dflow.sh).

  • SOC 2Coming soon
  • Australian DPACompliant
  • GDPRCompliant

The published trust center describes dFlow Cloud. This site is the new control plane; its security notes are on /security/ and /trust/.

08Live today vs in development

Live today. In development.

Live today

  • Deploys from an image, a GitHub repo, or a local directory
  • Releases and rollback
  • MCP and skills for agents
  • Postgres with on-demand backups
  • Volumes, jobs and cron jobs
  • Routes, custom domains, automatic TLS
  • Write-only encrypted secrets
  • Your own cluster, from the CLI
  • A Valkey cache
  • Restart the running pods without a new release
  • Google or GitHub sign-in, when the server configures it
  • Members on a workspace that is not personal

In development

  • Release botIN DEVELOPMENT
  • Alerts botIN DEVELOPMENT
  • Maintenance botIN DEVELOPMENT
  • One-click MCP connectIN DEVELOPMENT
  • The @dflow-sh/mcp npm packageIN DEVELOPMENT

We record each change in the public changelog. Read the changelog →

CLI-firstagent-safeyour own clusterdFlowRailwayCoolifyDokployHeroku

Get a beta invite.

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

Not wired yet. This preview has no backend, so nothing was sent.

Open to the first 100 users.

Book a design-partner call

Invite only. Read the beta terms.

FAQ

Questions

Is dFlow ready for production?

No, not yet. dFlow is in an invite-only beta. The product can change, and we offer no SLA. Read the beta terms first.

What does dFlow cost?

Pricing isn't set yet. The beta is invite only. The pricing page shows the beta limits for each workspace.

Do I need to know Kubernetes?

No. You deploy with the dflow CLI or your coding agent. dFlow creates the Kubernetes parts. You need a kubeconfig only for your own cluster.

Can my agent delete things?

Only with explicit confirmation. A database restore needs confirm_name. A route delete needs confirm_host. Other destructive actions need a confirmation argument too.

Can I use my own cluster?

Yes. Register it with dflow clusters:register. We store the kubeconfig encrypted. This works from the CLI only. An operator installs k3s or kubeadm with dflow-server cluster:create.

How is dFlow different from Railway or Coolify?

dFlow is CLI-first and agent-safe, on Kubernetes. Railway also offers agent tooling. Our focus is safe agent actions, rollback and your own cluster. See the compare pages.