Start deliberately vulnerable web apps, APIs and benchmark targets in Docker — from a terminal menu or one command — published on 127.0.0.1 by default.
- Features
- Safety model
- Quick start
- Usage
- Supported labs
- Installation
- Adding a lab
- Development and tests
- Web catalog
- Roadmap
- Disclaimer
- License
- Lab manager, not a launcher: create, start, stop, inspect, view logs and remove labs; several labs (and several copies of one lab) run side by side, each on its own port.
- Safe defaults: labs listen on
127.0.0.1only; exposing one to your network is an explicit, warned opt-in. - Only touches its own containers: everything WebVuln-Runner creates is labelled
io.webvuln-runner.managed=true; nothing else on the host is listed, stopped or removed. - Readiness checks: the container being "running" is not the same as the app being ready — WebVuln-Runner waits until the app answers and tells you when it does.
- Single-container and Docker Compose labs: multi-service labs such as OWASP crAPI and NodeGoat start as one unit.
- Beyond classic web labs: REST and GraphQL APIs, Node.js, Java, Python and PHP stacks, and the OWASP Benchmark for measuring SAST/DAST tools.
- One TUI for every lab: a terminal menu to add, start, stop, inspect and remove labs. The same actions are also available as commands (see Command line), which the tests use.
- Pick by what you want to learn: every lab is tagged with one or more of 10 vulnerability topics (injection, XSS & CSRF, access control, authentication, SSRF, files, API, secrets & config, dependencies, logic & LLM) and a level. Use
webvuln-runner catalog --topic ssrf --level beginner, the menu's "By topic" option, or the filters on the web catalog. - One catalog:
src/catalog.jsondrives the TUI, the commands, this README's lab table and the web catalog. It records each lab's license and the date it was last verified to start.
The labs are intentionally vulnerable. Treat every one of them as hostile code.
- Labs publish on
127.0.0.1. Use--bind 0.0.0.0(or the TUI's bind option) only on an isolated lab network. - Every lab runs on its own Docker network, so one lab cannot reach another. The integration test checks this by name and by IP.
- Lab containers run with
no-new-privileges, a 1024-process limit, a memory limit (at least 512 MB, or twice the lab's expected RAM) and at most 2 CPUs. A lab that genuinely breaks under these limits can opt out in the catalog ("hardening": false). - Compose labs must follow the same rules, and
scripts/check-compose-policy.shenforces them in CI. Every service must carry the WebVuln-Runner label andno-new-privileges. Ports must be published on${WVR_BIND}.container_name, privileged mode, host networking or PID namespaces, and Docker socket mounts are not allowed. This rule is why BrokenCrystals ships without its upstreamwatchtowerservice. - Docker isolates processes, but it is not a security boundary you should rely on against a compromised container. Do not run labs on machines holding sensitive data, and never on production infrastructure.
- Running Docker requires root, or membership in the
dockergroup — which is equivalent to root on that host. WebVuln-Runner itself does not need root; it needs access to the Docker daemon. - Some labs attack the client rather than wait to be attacked. DVMCP's challenges (tool poisoning, token theft and others) target the AI/MCP client that connects to it. Use a throwaway client with no access to real files, API keys or a shell.
- Images come from third parties (see each lab's source link). The catalog pins version tags where upstream publishes them; a weekly audit checks that every image can still be pulled.
# Download the installer of a tagged release, read it, then run it
curl -fsSLO https://raw.githubusercontent.com/yusufarbc/webvuln-runner/v2.3.1/installers/install.sh
less install.sh
sudo bash install.sh
# Start a lab and open the URL it prints
sudo webvuln-runner up juice-shopRun webvuln-runner without arguments. Use the arrow keys to move, Enter to choose, Tab to switch between buttons and Esc to go back.
Main menu. It lists your labs with their state and URL, followed by the actions: add a lab, stop all labs, remove all labs, and choose where new labs listen. Pick Quit to leave; your labs keep running.
Add a lab. Choose how to find one: the whole catalog, by vulnerability topic, or by level.
Picking by topic shows how many labs cover each topic:
Then choose a lab from the matching ones:
Accept or change the suggested name and host port (a free port is pre-filled), then confirm:
WebVuln-Runner pulls the image, starts the lab and waits until the app answers. Then it shows the URL to open. If the lab has default credentials or first-run steps, they are shown next.
Manage a lab. Select a lab in the main menu to start or stop it, check its status, read its recent logs or remove it. Removing asks twice and asks separately before deleting data volumes.
Status shows whether the app answers, its ports and its CPU and memory use. For a failed lab it also shows the exit code and the last log lines.
webvuln-runner catalog list labs that can be started
webvuln-runner list list your labs and their state
webvuln-runner up <lab> [options] create and start a lab
--name NAME instance name (default: wvr-<lab>)
--port PORT host port for the lab's main port (default: first free one)
--bind ADDR address to publish on (default: 127.0.0.1)
--no-wait do not wait for the app to answer
webvuln-runner start|stop|status <name>
webvuln-runner stop --all stop every lab
webvuln-runner stats [name] CPU and memory of running labs
webvuln-runner logs <name> [--tail N]
webvuln-runner rm <name> [--volumes]
webvuln-runner rm --all --yes [--volumes] remove every lab (never other containers)
webvuln-runner validate [catalog.json]
webvuln-runner uninstall
Example:
$ webvuln-runner up dvga
Lab 'wvr-dvga' (dvga) started.
web: http://127.0.0.1:5013/
Waiting for 'wvr-dvga' to answer (up to 90s)...
'wvr-dvga' is ready: http://127.0.0.1:5013/
$ webvuln-runner up dvga --name dvga-2 # a second copy gets the next free port
$ webvuln-runner list
NAME LAB STATE URL
dvga-2 dvga running http://127.0.0.1:5014/
wvr-dvga dvga running http://127.0.0.1:5013/Exit codes: 0 ok, 1 error, 2 usage error, 3 lab started but did not answer in time.
list and the menu show each lab's state:
runningorstopped.startingorunhealthy, from the image's own health check.restarting.partial n/m, when only some containers of a multi-container lab run.failed, when a container exited with an error code.statusthen shows the exit code and the last log lines.WVR_BINDchanges the default bind address, andWVR_HEALTH_TIMEOUTchanges the default readiness timeout in seconds.
| Lab | ID | Category | Level | Topics | Type | Default port(s) | License | Verified |
|---|---|---|---|---|---|---|---|---|
| OWASP Juice Shop | juice-shop |
web | beginner | injection, xss-csrf, access-control, authentication, ssrf, files, secrets-config, dependencies, logic-llm | image | 3000 | MIT | 2026-10-17 |
| DVWA | dvwa |
web | beginner | injection, xss-csrf, authentication, files | image | 8081 | GPL-3.0 | 2026-10-17 |
| OWASP WebGoat | webgoat |
training | beginner | injection, xss-csrf, access-control, authentication, ssrf, files, secrets-config, dependencies | image | 8080, 9090 | GPL-2.0-or-later | 2026-10-17 |
| bWAPP | bwapp |
web | beginner | injection, xss-csrf, authentication, ssrf, files, secrets-config, dependencies | image | 8082 | unknown | 2026-10-17 |
| OWASP Mutillidae II | mutillidae |
web | beginner | injection, xss-csrf, access-control, authentication, files, secrets-config | image | 8083 | GPL-3.0 | 2026-10-17 |
| XVWA | xvwa |
web | beginner | injection, xss-csrf, access-control, authentication, ssrf, files, secrets-config, dependencies | image | 8084 | GPL-3.0 | 2026-10-17 |
| VAmPI | vampi |
rest-api | beginner | injection, access-control, authentication, api | image | 5000 | MIT | 2026-10-17 |
| DVNA | dvna |
web | beginner | injection, xss-csrf, access-control, authentication, secrets-config, dependencies | image | 9091 | MIT | 2026-10-17 |
| VulnLab | vulnlab |
web | beginner | injection, xss-csrf, access-control, authentication, files, dependencies, logic-llm | image | 8085 | CC-BY-NC-SA-4.0 | 2026-10-17 |
| SQLi-Labs | sqli-labs |
web | beginner | injection | image | 8086 | unknown | 2026-10-17 |
| OWASP VulnerableApp | vulnerableapp |
web | intermediate | injection, xss-csrf, authentication, ssrf, files | image | 9092 | Apache-2.0 | 2026-10-17 |
| PyGoat | pygoat |
web | beginner | injection, xss-csrf, access-control, authentication, ssrf, secrets-config, dependencies | image | 8100 | MIT | 2026-10-17 |
| DVGA | dvga |
graphql | intermediate | injection, xss-csrf, access-control, authentication, ssrf, files, api, secrets-config | image | 5013 | MIT | 2026-10-17 |
| Hackazon | hackazon |
web | intermediate | injection, xss-csrf, files | image | 8087 | Apache-2.0 | 2026-10-17 |
| OWASP Benchmark | owasp-benchmark |
benchmark | advanced | injection, xss-csrf, files, secrets-config | image | 8443 | GPL-2.0 | 2026-10-17 |
| DVWS (PHP) | dvws |
rest-api | intermediate | injection, authentication, ssrf, secrets-config | image | 8088 | Apache-2.0 | 2026-10-17 |
| DSVPWA | dsvpwa |
web | beginner | injection, xss-csrf, authentication, files, secrets-config, dependencies | image | 65413 | MIT | 2026-10-17 |
| APIGoat | apigoat |
rest-api | beginner | access-control, authentication, ssrf, api, secrets-config, logic-llm | image | 8000, 8001-8010 | GPL-3.0 | 2026-10-18 |
| OWASP crAPI | crapi |
rest-api | intermediate | injection, access-control, authentication, ssrf, api, logic-llm | compose | 8888, 8025 | Apache-2.0 | 2026-10-17 |
| OWASP NodeGoat | nodegoat |
web | beginner | injection, xss-csrf, access-control, authentication, ssrf, secrets-config, dependencies | compose | 4000 | Apache-2.0 | 2026-10-17 |
| DVWS-Node | dvws-node |
rest-api | intermediate | injection, xss-csrf, access-control, authentication, ssrf, files, api, secrets-config, dependencies | compose | 8089 | GPL-3.0 | 2026-10-17 |
| Snyk Goof | goof |
web | beginner | injection, xss-csrf, secrets-config, dependencies | compose | 3001 | Apache-2.0 | 2026-10-17 |
| gRPC Goat | grpc-goat |
grpc | intermediate | injection, authentication, ssrf, api, secrets-config | compose | 50051, 50052-50058 | MIT | 2026-10-17 |
| DVMCP | dvmcp |
ai-mcp | intermediate | access-control, secrets-config, logic-llm | compose | 9001, 9002-9010 | unknown | 2026-10-17 |
| OWASP WrongSecrets | wrongsecrets |
training | beginner | secrets-config | image | 8090 | AGPL-3.0 | 2026-10-17 |
| BrokenCrystals | brokencrystals |
web | intermediate | injection, xss-csrf, access-control, authentication, ssrf, files, api, secrets-config, logic-llm | compose | 3100, 1080 | MIT | 2026-10-17 |
| DjanGoat | djangoat |
web | beginner | injection, xss-csrf, access-control, authentication, secrets-config | compose | 8110 | MIT | 2026-10-17 |
| OWASP Security Shepherd | security-shepherd |
training | beginner | injection, xss-csrf, access-control, authentication, secrets-config | compose | 8445 | GPL-3.0 | 2026-10-17 |
| WebSocket GOAT | wsgoat |
web | intermediate | access-control, authentication, xss-csrf, api | compose | 8096 | unknown | 2026-10-18 |
| VulnerableLightApp | vulnerablelightapp |
rest-api | intermediate | injection, access-control, ssrf, files, api, secrets-config, dependencies | compose | 8097 | GPL-3.0 | 2026-10-18 |
| Topic | Labs |
|---|---|
access-control |
juice-shop, webgoat, mutillidae, xvwa, vampi, dvna, vulnlab, pygoat, dvga, apigoat, crapi, nodegoat, dvws-node, dvmcp, brokencrystals, djangoat, security-shepherd, wsgoat, vulnerablelightapp |
api |
vampi, dvga, apigoat, crapi, dvws-node, grpc-goat, brokencrystals, wsgoat, vulnerablelightapp |
authentication |
juice-shop, dvwa, webgoat, bwapp, mutillidae, xvwa, vampi, dvna, vulnlab, vulnerableapp, pygoat, dvga, dvws, dsvpwa, apigoat, crapi, nodegoat, dvws-node, grpc-goat, brokencrystals, djangoat, security-shepherd, wsgoat |
dependencies |
juice-shop, webgoat, bwapp, xvwa, dvna, vulnlab, pygoat, dsvpwa, nodegoat, dvws-node, goof, vulnerablelightapp |
files |
juice-shop, dvwa, webgoat, bwapp, mutillidae, xvwa, vulnlab, vulnerableapp, dvga, hackazon, owasp-benchmark, dsvpwa, dvws-node, brokencrystals, vulnerablelightapp |
injection |
juice-shop, dvwa, webgoat, bwapp, mutillidae, xvwa, vampi, dvna, vulnlab, sqli-labs, vulnerableapp, pygoat, dvga, hackazon, owasp-benchmark, dvws, dsvpwa, crapi, nodegoat, dvws-node, goof, grpc-goat, brokencrystals, djangoat, security-shepherd, vulnerablelightapp |
logic-llm |
juice-shop, vulnlab, apigoat, crapi, dvmcp, brokencrystals |
secrets-config |
juice-shop, webgoat, bwapp, mutillidae, xvwa, dvna, pygoat, dvga, owasp-benchmark, dvws, dsvpwa, apigoat, nodegoat, dvws-node, goof, grpc-goat, dvmcp, wrongsecrets, brokencrystals, djangoat, security-shepherd, vulnerablelightapp |
ssrf |
juice-shop, webgoat, bwapp, xvwa, vulnerableapp, pygoat, dvga, dvws, apigoat, crapi, nodegoat, dvws-node, grpc-goat, brokencrystals, vulnerablelightapp |
xss-csrf |
juice-shop, dvwa, webgoat, bwapp, mutillidae, xvwa, dvna, vulnlab, vulnerableapp, pygoat, dvga, hackazon, owasp-benchmark, dsvpwa, nodegoat, dvws-node, goof, brokencrystals, djangoat, security-shepherd, wsgoat |
30 labs. Generated from src/catalog.json by scripts/gen-readme-table.sh. "Verified" is the last date scripts/labs-smoke.sh started the lab successfully.
Default credentials and first-run steps are shown after a lab starts. They are also listed in the catalog under credentials and notes.
installers/install.sh supports Debian / Ubuntu / Kali, Fedora / RHEL / CentOS / Rocky / AlmaLinux and Arch Linux. It:
- installs Docker (from Docker's own repository on Red Hat family systems), the Compose v2 plugin where available,
whiptail,jqandcurl, and enables the Docker service; - downloads the release tarball, verifies it against the release's
SHA256SUMSand refuses to install on mismatch. Every tarball also carries a signed build provenance attestation, so you can check that it was built by this repository's release workflow and not uploaded by hand:gh release download v2.3.1 --repo yusufarbc/webvuln-runner --pattern '*.tar.gz' gh attestation verify webvuln-runner-2.3.1.tar.gz --repo yusufarbc/webvuln-runner - installs to
/usr/local/lib/webvuln-runner, withwebvuln-runner(and the 1.x nameWebVuln-Runner) in/usr/local/bin.
Options are environment variables: WVR_RELEASE=v2.3.1, WVR_DISTRO=debian|redhat|arch, WVR_SKIP_DEPS=1.
From a git checkout: sudo installers/install-files.sh installs the files of the checkout. You can also run src/webvuln-runner directly without installing.
Update: run the installer of the newer release; your labs are kept. See CHANGELOG.md for what changed.
Rollback: if a release misbehaves, install the previous one over it. Labs and their volumes are kept:
curl -fsSLO https://raw.githubusercontent.com/yusufarbc/webvuln-runner/v2.2.0/installers/install.sh
sudo WVR_RELEASE=v2.2.0 bash install.shUninstall: sudo webvuln-runner uninstall. It asks before removing the labs it created and never touches other containers. Pass --remove-labs or --keep-labs to skip the question when calling /usr/local/lib/webvuln-runner/uninstall.sh directly.
Upgrading from 1.x: the installer removes the old /bin/WebVuln-Runner and /bin/indexApp.lst. Containers created by 1.x carry no WebVuln-Runner label, so they do not appear in the new menu. Remove them with docker rm -f <name>.
Without the Compose plugin, multi-container labs are shown as unavailable and everything else works.
Add an entry to src/catalog.json:
{
"id": "my-lab",
"name": "My Lab",
"description": "One line about what it teaches",
"category": "web",
"stack": ["PHP", "MySQL"],
"type": "image",
"image": "vendor/my-lab:1.2.3",
"ports": [{ "name": "web", "container": 80, "host": 8099, "scheme": "http", "path": "/" }],
"ram_mb": 256,
"credentials": "admin / admin",
"topics": ["sqli", "xss", "csrf"],
"level": "beginner",
"license": "MIT",
"verified": "2026-10-17",
"source": "https://github.com/vendor/my-lab"
}| Field | Meaning |
|---|---|
category |
web, rest-api, graphql, grpc, ai-mcp, benchmark or training |
topics |
what the lab lets you practise, from: injection (SQL/NoSQL, command, SSTI, XXE), xss-csrf, access-control (IDOR/BOLA, authorization), authentication (login, sessions, JWT), ssrf, files (path traversal, upload), api (API Top 10, GraphQL, gRPC), secrets-config (secrets, misconfiguration, crypto), dependencies (vulnerable components, deserialization), logic-llm (business logic, LLM/MCP). Take them from the lab's own documentation |
level |
beginner, intermediate or advanced |
license |
the upstream project's SPDX license id, or unknown |
verified |
date the lab last started successfully (scripts/labs-smoke.sh --update-verified <id> sets it) |
type |
image (one container) or compose (a Compose project, see below) |
image |
repository:tag. Use a version tag whenever upstream publishes one |
ports |
first entry is the main port (shown as the URL and used for the readiness check); host defaults must be unique across the catalog |
scheme |
http, https, tcp or grpc |
health |
optional { "type": "http" | "https" | "tcp", "path": "/ready" } override |
health_timeout |
seconds to wait for readiness (default 90) |
env, workdir, command |
optional container environment, working directory and command |
mem_limit_mb, cpus |
optional resource limits (defaults: max(512, 2 × ram_mb) MB, 2 CPUs) |
hardening |
false disables no-new-privileges and the limits; only for labs that break, with the reason in notes |
Compose labs set "type": "compose" and "compose": "compose/<id>/compose.yml". In that file:
- every service carries the
x-wvr-labelsblock (seesrc/compose/nodegoat/compose.yml); - every published port is written as
"${WVR_BIND:-127.0.0.1}:${WVR_PORT_<NAME>:-<default>}:<container port>", where<NAME>is the upper-cased port name from the catalog; container_nameis not used, so several instances can coexist;- every service sets
security_opt: ["no-new-privileges:true"](scripts/check-compose-policy.shchecks all of this).
Then run src/webvuln-runner validate, scripts/gen-readme-table.sh and start the lab once with src/webvuln-runner up <id>.
src/webvuln-runner validate # catalog schema checks
bats tests/unit # unit tests with a fake docker CLI
bash tests/integration.sh # end-to-end against a real Docker daemon
scripts/check-compose-policy.sh # security rules for every Compose lab
scripts/ci-local.sh # all of the CI checks above in one go
scripts/labs-smoke.sh [--report-md] [lab ...] # start lab(s) for real, with JSON/MD reporting
shellcheck -x src/webvuln-runner src/lib/*.sh installers/*.sh installers/*/*.sh scripts/*.shCI runs all of these on every pull request. The integration test checks that labs bind to 127.0.0.1 and that busy ports are skipped. It also checks that a container not created by WebVuln-Runner survives every command.
index.html (webvulnrunner.dev) renders the catalog with search and category filters. It also shows the command that starts each lab. It is static and read-only on purpose: a browser page cannot safely control your Docker daemon, and exposing the Docker socket to one would hand over root on your machine.
WebVuln-Runner runs labs; it does not scan, test or modify them. The labs' code, images and content belong to their upstream projects. The tool only decides how they run: on loopback, on their own network, with resource limits, and without services that would endanger the host.
- Pin images that only publish
latestby digest, and report when upstream changes them webvuln-runner reset <name>to restart a lab with clean datawebvuln-runner updateto pull newer catalog images- An advanced profile for Kubernetes labs (e.g. Kubernetes Goat), which need a cluster rather than Docker alone
- Optional local API for the web catalog, with authentication
See CONTRIBUTING.md for full guidelines on adding new labs, Compose policies, and running the local test suite.
- Create a branch:
git checkout -b feature/my-lab - Add the lab as described in Adding a lab and run
./scripts/ci-local.sh - Open a pull request describing the lab and how you tested it
Report bugs or request labs at https://github.com/yusufarbc/webvuln-runner/issues.
These applications are intentionally vulnerable. Do not expose them to the internet or run them on production infrastructure. Use them only on systems you own or are authorized to test. See SECURITY.md for the project security policy, threat model, and responsible disclosure procedures.
GNU General Public License v3.0 — see LICENSE. Third-party lab images and the adapted Compose files keep their own licenses, noted at the top of each file under src/compose/.







