MCP server included · 34 agent tools

Deploy from anywhere. Even from a chat.Even from the beach.

Ask Claude, Cursor or Codex. Run one command. Or just git push. DeployCloud builds your repo and puts it live on HTTPS — any language, zero downtime, no servers to run.

Start freeWatch an agent ship it
  • Free tier
  • Any language
  • No credit card

🏖️ Beach mode: deploy from the beach — your agent watches prod while you watch the waves.

🌌 Built for the 2am deploy — from bed, from a train, from orbit.

Ask. Approve. Live.

The whole setup, from your phone if you like — you approve, the agent does everything else.

1 · Ask your agent

Deploy this repo on DeployCloud — docs: deploycloud.app/mcp.mjs

On it — fetching the MCP server…

That link IS the MCP server. The agent grabs the CLI from /cli.mjs and runs deploycloud login — no config written by hand.

2 · Approve on your phone

Authorize the CLI

FJK3-9QXP

claude@your-machine wants access

Authorize

Open the agent’s link anywhere, sign in (or create your account), tap Authorize. The token mints itself — nobody copies secrets around.

3 · It ships

→ building with Nixpacks…

→ releasing web.1

✓ health check passed

https://myapp.deploycloud.app

The agent deploys, watches the build, and sends you the live HTTPS URL. From then on it’s just asking.

Watch a whole session

Four assistants, four jobs — deploy, roll back, scale, wire a database. Real tool calls, real arguments.

Deploy from anywhere

Three ways to ship. One platform.

Whatever you have to hand — an AI assistant, a terminal, or a git remote — is a complete deploy pipeline. Same API underneath, same result.

01 · MCP

Ask an agent

Point Claude, Cursor, Codex or any MCP client at DeployCloud and it gets 34 real tools: deploy, logs, rollback, env vars, add-ons, domains and crons.

$ claude mcp add deploycloud \ -e DEPLOYCLOUD_TOKEN=dc_… \ -- deploycloud-mcp“roll api back to v47”✓ rolled back · v47 live
02 · CLI

Run one command

One zero-dependency script and an API token. deploy --watch streams the build and exits on success or failure, so it drops straight into CI.

$ deploycloud deploy api --watch[build] detected Dockerfile[release] health check passedok deployment #42 is running$ deploycloud scale api web=3ok api scaled to web=3
03 · Git

Just push

Connect GitHub and every push to your branch builds and releases itself. A Procfile is all the configuration there is.

# Procfileweb: node server.jsworker: node worker.js$ git push deploycloud main→ Building with Nixpacks…✓ Live at myapp.deploycloud.app

Connect your agent in three steps

  1. 1Create an API token — it’s shown once, so copy it right away.
  2. 2Copy the message on the right into Claude, Cursor, Codex — any MCP client — with your token pasted in.
  3. 3Ask for a deploy. The agent downloads the server, connects, and ships — you never leave the chat.
message for your assistant
Connect DeployCloud (my PaaS) over MCP, then deploy my app.

1. Download the MCP server — one zero-dependency Node 20+ script:
     curl -fsSL https://deploycloud.app/mcp.mjs -o ~/.deploycloud-mcp.mjs

2. Register it with yourself, with these env vars:
     DEPLOYCLOUD_URL=https://deploycloud.app
     DEPLOYCLOUD_TOKEN=<my token>
   Claude Code:
     claude mcp add deploycloud \
       -e DEPLOYCLOUD_URL=https://deploycloud.app \
       -e DEPLOYCLOUD_TOKEN=<my token> \
       -- node ~/.deploycloud-mcp.mjs
   Cursor / other MCP clients: add an mcpServers entry with command
   "node", args ["<absolute home>/.deploycloud-mcp.mjs"], and those env vars.

3. Verify with list_apps. Then create_app and deploy_app for my repo —
   or upload_bundle if I point you at a local directory — and poll
   get_deployment until it reaches "running". Give me the live URL.

My token (create at https://deploycloud.app/tokens — shown once): <PASTE TOKEN>

No token pasted above? Mint one yourself: fetch https://deploycloud.app/cli.mjs
and run `node cli.mjs login https://deploycloud.app` — I'll approve the link
it prints, and the config lands where the MCP server reads it.

No laptop? Still fine.

Your assistant runs the MCP server wherever it already lives — desktop app, terminal, editor. From a phone you go at the platform directly instead: every operation is a plain HTTPS call with a bearer token, and the dashboard is a responsive web app. Deploy, roll back and read logs from the train.

See the 34 toolsStart free
apirunning

v48 · main @ 4e91c2f

deployed 2m ago

DeployRollback

web.1 GET /healthz 200

web.1 GET /api/orders 200

Everything you need to ship

A complete, fully managed platform — no per-seat pricing, no infrastructure to run.

Any language

Bring a Dockerfile or let Nixpacks auto-detect Node, Python, Go, Ruby, PHP, Rust, Java and more.

Zero-downtime deploys

Health-checked rolling releases behind Traefik. New version goes live only once it passes.

Preview deployments

Every branch gets its own live URL. Previews post a GitHub commit status and a PR comment — and can sit behind a password.

Scale & autoscale

Scale processes horizontally with one click, or let app-level autoscaling track CPU. Grow onto a Swarm fleet with on-demand VPS nodes.

One-click add-ons

Provision Postgres, Redis and MinIO in a click. Connection strings wired into your env for you.

Rollback & log drains

One-click rollback to any release, streaming container logs, and log drains that forward them to an HTTP endpoint or syslog.

And everything else you stop noticing once it just works:

Persistent volumesCron jobsProcfile processesMetrics & alertsScheduled Postgres backupsOne-click restoreCustom domainsAutomatic HTTPSMaintenance modeCached buildsActivity audit trailScoped env vars

Built for the whole lifecycle

Not just the first deploy — reviewing, configuring, watching and recovering, all in one place.

Preview deployments

Every branch gets a real URL

Push a branch and DeployCloud builds it into its own live environment on its own subdomain — production stays untouched. Share the link in the pull request, click around the actual app, then merge with confidence.

  • Built automatically on push, torn down when the branch is deleted
  • Isolated containers and env — previews can never touch production
  • Same build pipeline as production, so what you review is what ships
myapp — previews

  • main

    https://myapp.deploycloud.app

    production

  • feat/checkout

    https://myapp-pv-feat-checkout…

    live

  • fix/login-loop

    https://myapp-pv-fix-login-loop…

    building

Rollback & observability

Bad deploy? Un-ship it in one click

Every release is kept, image and all, so rolling back is instant — no rebuild, no waiting. Live CPU and memory stats, streaming container logs and a full activity feed keep you certain about what is running and why.

  • One-click rollback to any previous release, images retained
  • A release only goes live after it passes its health check
  • Audit trail of every deploy, scale, env and add-on change
myapp — releases
  • v48

    deploy · main @ 4e91c2f

    live
  • v47

    rollback → v45

    supersededRollback
  • v46

    deploy · main @ b7d01aa

    failed health checkRollback

web.1 GET /api/orders 200 12ms

worker.1 processed job email:4192

Live directory

Real apps, running right now

Not logos on a wall. Every app in the showcase is deployed on this platform this minute — with its live status, replica count, a screenshot of its actual homepage and a link you can click. Owners opt in per app; nothing is listed without them.

Browse live appsPut yours here
notes-app
metrics-hub
storefront.deploycloud.app
storefrontrunning

2 replicas · deployed 4m ago

How it works

From a Git repo to a production URL in three steps.

  1. 1

    Point us at a Git repo

    Connect any repository and pick a branch — via GitHub push-to-deploy, the CLI, or the REST API.

  2. 2

    We build & release

    We build with your Dockerfile or Nixpacks, then release behind Traefik with automatic HTTPS.

  3. 3

    Your app is live

    It goes live on its own subdomain, health-checked and ready. Scale it out with one click.

The middle path you were looking for

Big-cloud simplicity without the per-seat bill. Full control without a platform team.

DeployCloudBig-cloud PaaSDIY / self-hosted
Deploy from an AI agent (MCP)
34 tools, built in
no MCP server
build it yourself
Zero servers to run or patch
you own the ops
Zero-downtime health-checked releases
if you wire it up
Preview environment per branch
often a paid tier
Databases, cache & storage add-ons
metered per add-on
operate them yourself
Built-in metrics & threshold alerts
often a paid tier
run Prometheus yourself
Scheduled backups & one-click restore
metered / paid tier
script and babysit it
Flat, predictable pricing
one monthly plan
per app, per seat, per GB
cheap infra, costly ops
See how we compare to:RailwayRenderDokkuCoolifyFly.io

Simple, flat pricing

Start free. Upgrade to a flat monthly plan — no per-seat pricing, no metered add-ons, no surprise egress fees.

Free

Kick the tires with a single app.

$0/mo
Start free
  • 1 app
  • 512 MB / 0.5 vCPU
  • Automatic HTTPS
  • Preview deploys
Most popular

Pro

For side projects that outgrew free.

$19/mo
Choose Pro
  • 10 apps
  • 1 GB / 1 vCPU each
  • Up to 3 replicas
  • Add-ons (Postgres, Redis, S3)
  • Backups (7-day retention)
  • Alerts & deploy notifications

Business

Production workloads and teams.

$99/mo
Choose Business
  • 50 apps
  • 2 GB / 2 vCPU each
  • Up to 10 replicas
  • Backups (30-day retention)
  • Autoscaling
  • Log drains
  • Priority everything

No credit card to start · cancel anytime · every plan includes automatic HTTPS, preview deploys and zero-downtime releases.

Frequently asked questions

Short answers to the things people ask before their first deploy.

Can an AI agent deploy for me?

Yes. DeployCloud ships an MCP (Model Context Protocol) server with 34 tools, so Claude, Cursor or any MCP client can deploy, tail logs, roll back, set environment variables, run one-off commands and manage add-ons, domains and cron jobs. It calls the same API the CLI does — it is not a separate “AI mode” with different rules.

Can I deploy from my phone?

Yes. Every operation is a plain HTTPS API call authenticated with a bearer token, and the dashboard is a normal web app — so deploying, rolling back and reading logs all work from a phone browser. (The MCP server itself speaks stdio, so it runs wherever your AI assistant runs, not on the phone.)

What do I need to get started?

A Git repo. Connect GitHub (or push any repository) and DeployCloud builds and runs it for you — no servers to provision, no Docker to install, no proxy to configure. HTTPS, subdomains and TLS certificates are issued automatically. You can be live in minutes on the free tier.

Which languages and frameworks are supported?

All of them. If your repo has a Dockerfile we build exactly that. If it doesn’t, Nixpacks auto-detects Node, Python, Go, Ruby, PHP, Rust, Java and more, and produces the image for you. If it runs in a container, it runs on DeployCloud.

Do preview deployments touch production?

Not its infrastructure. Each branch builds into its own containers on its own subdomain with its own lifecycle, gets its own empty persistent volumes, and never publishes production’s custom domains — so a branch cannot touch production’s disk or take its traffic. One thing IS shared on purpose: environment variables scoped to “all”, which includes add-on credentials — so a preview talks to the same database unless you scope that variable to production. Previews are torn down automatically when the branch is deleted.

How are secrets and env vars handled?

App environment variables and add-on credentials are encrypted at rest with AES-256-GCM. They are decrypted only to inject into your app at runtime, and are never shown in build logs, deploy output or the dashboard once saved.

Can my app scale?

Yes. Scale out horizontally with one click, or turn on app-level autoscaling and DeployCloud adds and removes instances automatically to track CPU as traffic changes. Zero-downtime, health-checked releases mean scaling never drops a request.

Does it have monitoring and alerting?

Every app gets a time-series chart of CPU and memory sampled by the worker, on top of live stats and streaming logs. Set a CPU or memory threshold and DeployCloud pings your webhook when it is crossed, and again when it recovers. You can also drain runtime logs to any HTTP endpoint or syslog server.

Can I back up my databases?

Yes. Point-in-time is a cron away: schedule Postgres add-on backups with automatic retention and rotation, download any dump, and restore one back into your database with a single click.

What does it cost?

Start free — one app, automatic HTTPS and preview deploys, no credit card. Paid plans are a flat monthly price: Pro at $19/mo (10 apps, add-ons) and Business at $99/mo (50 apps, more resources). No per-seat pricing, no metered add-ons, no surprise egress fees — you always know your bill.

Deploy your first app today.

Point us at a repo and go — free to start, no credit card, cancel anytime.

Get startedSee pricing