docker-multitor

May 23, 2026 ยท View on GitHub

Dockerized multitor. Runs multiple Tor circuits behind HAProxy for load-balanced anonymous proxying.

Alpine 3.21 / Tor / Privoxy / HAProxy

usage

docker build -t multitor .
docker run --rm -p 16379:16379 multitor

curl --proxy http://localhost:16379 https://httpbin.org/ip

More Tor instances:

docker run --rm -p 16379:16379 -e TOR_INSTANCES=10 multitor

Or pull from registry:

docker run --rm -p 16379:16379 evait/multitor

env

vardefaultwhat it does
TOR_INSTANCES5number of Tor circuits to spawn

mirror

Robust parallel Tor mirror tool for downloading directory-listed URLs from .onion sites. Built for bulk-downloading leaked datasets for DSGVO analysis (informing affected users about stolen data).

Architecture: Python + SQLite job queue + parallel wget workers. Spider crawls directory listings, workers download files concurrently through multiple Tor circuits.

  • SQLite database tracks all files - crash-safe, resumable
  • Parallel download workers (one per Tor circuit)
  • Real-time terminal UI showing progress, speed, and worker status
  • Automatic infinite retry on network failure
  • Resume support - re-run the same command to pick up where you left off

install

Requires uv and wget.

# Install as global tool
uv tool install ./mirror

# Or run directly from repo
uv run --directory mirror mirror --help

# or install it from the online repo
uv tool install git+https://github.com/evait-security/docker-multitor/tree/master/mirror

usage

# Start multitor proxy
docker run --rm -d -p 16379:16379 -e TOR_INSTANCES=5 evait/multitor

# Mirror a .onion directory listing
mirror http://exampleonion.onion/files/

# Mirror to a specific destination with 10 workers
mirror --parallel 10 http://exampleonion.onion/dump/ /data/leak-mirror

# Custom proxy address
mirror --proxy http://10.0.0.5:16379 http://exampleonion.onion/files/

# Resume an interrupted download (just re-run the same command)
mirror http://exampleonion.onion/dump/ /data/leak-mirror

# Start fresh, ignoring previous state
mirror --fresh http://exampleonion.onion/dump/ /data/leak-mirror

options

flagdefaultdescription
url-URL to mirror (directory listing with links)
destination.Local directory to save files
-p, --parallel5Number of parallel download workers
--proxyhttp://127.0.0.1:16379Multitor proxy address
--timeout120Download timeout per file (seconds)
--fresh-Delete existing database, start over
-v, --verbose-Write detailed logs to mirror.log

how it works

  1. Spider crawls directory listings recursively via the Tor proxy, discovers all file URLs, stores them in a SQLite queue (.mirror.db in the destination)
  2. Workers (N parallel threads) pull jobs from the queue and download files using wget through the HAProxy round-robin - each request exits through a different Tor circuit
  3. Retry loop - failed downloads are requeued and retried indefinitely until all files are saved
  4. Resume - the database persists in the destination dir; re-running detects it and picks up where it left off

test

./test.sh

Builds the image, spins up a container, waits for Tor bootstrap, runs HTTP/HTTPS requests through the proxy, verifies Tor exit and round-robin. Cleans up after itself.

publish

./publish.sh 1.0.0

Tags and pushes to Docker Hub as evait/multitor:1.0.0 and evait/multitor:latest.

how it works

multitor spawns N Tor processes with separate SOCKS ports, puts a Privoxy HTTP proxy in front of each, then HAProxy round-robins across all Privoxy backends on port 16379. Each request potentially exits through a different Tor circuit.

credits