
A CTF platform with Docker-based labs is one that can start a live container for a challenge, such as a vulnerable web application or network service, and hand each participant an address to attack. The platforms differ in how they provide this. CTF-Hub and GZCTF include it as a native feature. CTFd provides it on its hosted tiers and in CTFd Enterprise, while self-hosters of open-source CTFd rely on third-party plugins such as ctfd-whale. rCTF integrates with a separate per-team instancer, and kCTF is a dedicated infrastructure project that you pair with a scoreboard.
Disclosure: this guide is written by the CTF-Hub team, and CTF-Hub is one of the options in it. We assessed every product from its vendor's public documentation, as of November 2026. "Not documented" means we found nothing on the pages we read. It does not mean the feature is absent.
Which CTF platforms provide Docker-based challenge instances?
| Platform | How challenge instances are provided | Instance models documented | Who runs the container hosts | Also documented |
|---|---|---|---|---|
| CTF-Hub | Native feature, from the Team plan | Shared on Team and above; per team on Pro and above; per participant on Enterprise | The vendor | Per-instance dynamic flags from Pro; VPN access from Pro; automatic cleanup |
| CTFd (hosted tiers) | Official hosted-tier feature: Docker images pushed to the CTFd registry run on container servers | Shared: a deployed service can be reached by any user who knows its address. Per-account instances are a separate feature the docs call Application Target | The vendor | Each hosted plan includes a stated number of container servers; linux/amd64 images only |
| CTFd Enterprise | Official Enterprise feature | Per participant: the vendor states each participant gets their own isolated infrastructure | The vendor on CTFd Cloud; you when self-hosted | Browser desktop workspaces |
| CTFd (open source) | Third-party plugin, for example ctfd-whale; no built-in instance launcher is documented | Depends on the plugin | You | Automatic deployment is limited to hosted CTFd and Enterprise installations |
| rCTF (OtterSec) | Optional integration with a separate instancer | Per team | You | Docker and Kubernetes backends |
| GZCTF | Native feature | Dynamic containers distributed and managed by the platform | You | Per-team dynamic flags; Docker or Kubernetes; traffic forwarding with optional capture |
| kCTF (Google) | Separate infrastructure project; it is not a scoreboard | A new sandboxed environment for every TCP connection | You, on your own Kubernetes cluster | A healthcheck container per challenge; optional proof-of-work |
| Hack The Box (CTF platform) | Vendor-managed, for Hack The Box content | Not documented in the pages we read | The vendor | Pwnbox, a browser attack machine, as an optional per-user add-on |
| RootTheBox | Not documented | Not documented | You | Docker is documented as a way to deploy the scoreboard itself |
| Mellivora | Not documented | Not documented | You | A docker-compose option for deploying the engine itself |
What are container-based challenge instances?
A container-based challenge instance is a running copy of a challenge's target system, packaged as a container image and started on demand by the CTF platform. The author builds the image once. When a participant opens the challenge, the platform starts a container from it, publishes an address, and stops the container when the time limit ends or the participant is finished.
Organizers want this for four reasons.
- A real target. Web, binary-exploitation and network challenges need something to attack. A downloadable file cannot stand in for a live service.
- No shared-state spoilers. On a shared target, one player's uploaded web shell, changed password or leftover file is visible to the next. With private instances, nobody inherits another player's progress.
- Clean reset. A participant who breaks their target can throw it away and start a fresh one, without asking an organizer to restart a server everyone depends on.
- Dynamic per-instance flags. Each instance can carry its own flag. A flag copied from another team does not validate, which discourages flag sharing and tells you who shared it.
What is the difference between shared, per-team and per-participant instances?
One shared instance for everyone. A single container serves the whole event. It is the cheapest model and the simplest to run. It works for challenges where players cannot change the target, such as a read-only web application. It fails when one player can crash or alter the service for the rest.
One instance per team. Each team gets a private copy that its members share. This matches how team competitions are scored, keeps teams from interfering with each other, and makes per-team dynamic flags possible. Capacity grows with the number of teams, not the number of people.
One instance per participant. Every person gets a private copy. This is the model for training and assessment, where you want to know what each individual did, and for events scored individually. It needs the most capacity.
kCTF takes a fourth approach for network services: its sandbox forks a new environment for every TCP connection, so isolation happens per connection instead of per account.
What does it take to run Docker-based labs in a real event?
Launching a container is the easy part. These five things decide whether instanced challenges work on the day, whether you run them yourself or ask a vendor how they handle them.
Isolation between players
Participants are running exploits against these containers, and some will gain code execution inside them. A container that can reach other players' containers, the host or the scoreboard is a problem. Check for network separation between instances, resource limits per container, and what an attacker who escapes the application can see.
Cleanup of abandoned instances
People start an instance, get stuck and close the tab. Without a time limit and automatic removal, idle containers pile up until the hosts are full and nobody else can start one. Look for an expiry timer, a limit on concurrent instances per team or person, and cleanup that does not depend on the player.
Capacity at event start
Demand peaks in the first minutes, when everyone opens their first challenge at once. Estimate the peak as teams or participants multiplied by the instances each may hold, then make sure the hosts and the image pulls can meet it. Large images that must be downloaded on first launch are a common cause of slow starts.
Network access: public endpoint or VPN
An instance can be published on a public address, which needs no setup from the player but exposes the target to the internet, or reached through a VPN, which keeps targets private and allows challenges that involve several machines or non-web protocols. VPN access adds a connection step that some participants will need help with.
Cost
Instanced challenges are the part of a CTF that scales with headcount. A shared instance costs the same for ten players or a thousand. Per-team and per-participant instances multiply compute by the number of teams or people, for as long as each instance lives. On hosted platforms this shows up as plan limits or tiers. On your own infrastructure it shows up on the cloud bill.
How does each CTF platform provide Docker-based labs?
CTF-Hub is a managed, multi-tenant cloud platform for hosting CTF competitions and hands-on security training. Docker container challenges are a native feature from the Team plan: isolated container environments launched per challenge, with automatic cleanup. Shared instances are available on Team and above, per-team instances on Pro and above, and per-participant instances on Enterprise. Per-instance dynamic flags start on Pro. Instances are reached over the public internet, or over VPN from Pro. It suits organizations that want instanced challenges without running container hosts. It is vendor-hosted only, not open source, and a young product (2026) with a smaller community than CTFd. Plans are on the pricing page.
CTFd (hosted tiers) provides containers as an official hosted feature. You push Docker images to CTFd's registry and they run on container servers; each CTFd hosted plan includes a stated number of them. The documentation is specific about the model: an application deployed this way can be accessed by any user who knows its URL or hostname, so it is a shared instance. Per-account instances are a separate feature the documentation calls "Application Target". Hosted deployment supports linux/amd64 images only. It suits organizers who want CTFd without operating it.
CTFd Enterprise is the tier where CTFd states that each participant gets their own isolated infrastructure. It also includes browser desktop workspaces, which the documentation says are currently available only in Enterprise. It is sold through sales and can be managed on CTFd Cloud or self-hosted.
CTFd (open source) has no built-in per-user instance launcher described in its README. The CTFd documentation says automatic deployment is limited to hosted CTFd and Enterprise installations. Self-hosters use third-party plugins, for example ctfd-whale, an MIT-licensed community project that is not a CTFd product. It suits teams with the skills to run and maintain both CTFd and a plugin.
rCTF separates the scoreboard from challenge hosting. The original project stated that rCTF itself does not handle challenge deployments. The maintained OtterSec version documents an optional integration with a per-team instancer, with Docker and Kubernetes backends. It suits CTF teams running public competitions on their own infrastructure.
GZCTF is an open-source CTF platform with native dynamic containers: it distributes and manages them on Docker or Kubernetes, with several port mapping methods, per-team dynamic flags, and traffic forwarding with optional capture. It suits self-hosting organizers who want instancing and dynamic flags in the platform itself. Its licence is an AGPL-3.0 core with additional restricted components.
kCTF is a separate infrastructure project, not a CTF platform. kCTF is CTF infrastructure from Google written on top of Kubernetes for managing, deploying and sandboxing challenges. Each challenge is deployed with a healthcheck container, sandboxed with nsjail, and can sit behind an optional proof-of-work. It has no scoreboard, so you pair it with one such as CTFd or rCTF. It suits organizers with Kubernetes skills who need hardened hosting for remote-exploitation challenges.
Hack The Box runs challenge infrastructure for you on its CTF platform, for events built from Hack The Box's own content. Pwnbox, a browser attack machine, is an optional per-user add-on. It suits organizations that want a ready-made event and do not need to deploy their own containers.
RootTheBox and Mellivora are open-source scoreboards for which we found no documented per-player instance launcher. Both document Docker as a way to deploy the scoreboard itself. Mellivora is not archived but has had no commits since 2023. If you need per-player instances with either, check with the project before relying on it.
Which option fits your event?
- A small event with mostly read-only web challenges: shared instances are enough. CTFd hosted container servers or CTF-Hub on the Team plan both cover this.
- A public team competition, run by people who like infrastructure: rCTF with its instancer, GZCTF, or open-source CTFd with a community plugin. These cost nothing to license and give you full control.
- A public competition with heavy binary-exploitation challenges: kCTF for the challenges, with CTFd or rCTF as the scoreboard.
- A team competition without an operations team: CTF-Hub on Pro, which adds per-team instances, dynamic flags and VPN access.
- Individual training or assessment, one environment per person: CTFd Enterprise or CTF-Hub Enterprise.
- An event on ready-made content: Hack The Box, which runs the infrastructure and supplies the challenges.
- An isolated network or strict data-residency rules: a self-hosted option. CTF-Hub cannot be self-hosted.
For the hosting question behind this one, see cloud-based CTF platforms and managed versus self-hosted CTF. For a broader feature checklist, see essential CTF platform features, and for other options near CTFd, CTFd alternatives.
Frequently Asked Questions
What is a Docker-based lab in a CTF?
A Docker-based lab is a challenge that runs as a live container, such as a web application or network service, which participants attack to find the flag. The platform starts the container, gives the participant an address to connect to, and removes it afterwards.
Which CTF platforms launch challenge containers natively?
As of November 2026, CTF-Hub and GZCTF document container instances as a built-in feature, and CTFd provides them on its hosted tiers and in CTFd Enterprise. rCTF integrates with a separate per-team instancer, and kCTF is dedicated challenge infrastructure that you pair with a scoreboard.
Does open-source CTFd launch a container per player?
Not by itself. The CTFd documentation limits automatic challenge deployment to hosted CTFd and Enterprise installations. Self-hosters of the open-source version use third-party community plugins such as ctfd-whale.
What is the difference between shared, per-team and per-participant instances?
A shared instance is one container that every participant attacks. Per-team instances give each team a private copy, and per-participant instances give each person a private copy. Private copies prevent players from interfering with each other but need more capacity.
What are dynamic flags?
A dynamic flag is a flag generated separately for each instance, so every team or participant has a different correct answer. A flag copied from another team does not validate, which discourages flag sharing and shows the organizer where a shared flag came from.
Which CTF-Hub plans include container challenges?
Container challenges are available from the Team plan. Shared instances are on Team and above, per-team instances on Pro and above, and per-participant instances on Enterprise. Dynamic flags and VPN access start on Pro.
Share this article
IIYour move
Ready to Launch Your CTF Competition?
Join thousands of organizations using CTF-Hub to train cybersecurity professionals worldwide.