Installation

One installer, one command. scripts/setup.sh is idempotent — running it twice never breaks anything.

TL;DR


git clone https://github.com/jacksonfdam/medusa-nexus.git
cd medusa-nexus
scripts/setup.sh
source ~/.mnexus/env.sh
mnexus doctor    # confirms every engine landed
mnexus           # opens the interactive REPL

On a fresh laptop this takes ~10 minutes (Ghidra and MobSF Docker dominate the wall clock).

Install modes

The setup script reads flags from scripts/setup.sh --help:

Flag	What it does
(no flag)	Full install: venv + every engine + Ghidra + MobSF + frida-server.
`--minimal`	Skip Ghidra, MobSF, and the frida-server push. Use for CI or quick eval.
`--device`	Only push the matching `frida-server` to the connected ADB device.
`--moxy`	Start Moxy in Docker, extract the mitmproxy CA, push it to the device, and write `MNEXUS_MOXY_*` to env.
`--ios-tools`	Install bagbak, ldid, and frida-ios-dump in one shot for iOS work.
`--doctor`	Skip install, just run `mnexus doctor`.

You can combine flags — e.g. scripts/setup.sh --minimal --moxy --ios-tools gives you a lean install with iOS + Moxy ready.

What gets installed where


$HOME/.mnexus/
├── tools/          ← Ghidra, jadx, frida-server, ldid, bagbak…
├── workspace/      ← per-project artefacts (apks, decompiled trees, reports)
├── nexus.sqlite3   ← single-file artifact store
└── env.sh          ← source this from your shell rc

The installer never writes into /usr/local/bin or your $PATH directly. Everything lives under $MNEXUS_HOME (default ~/.mnexus) and gets surfaced via env.sh.

Step by step

macOS


# 1. Toolchain prerequisites — Homebrew handles the rest.
xcode-select --install
brew install python@3.13 docker openjdk@17 adb
 
# 2. Clone + run the installer.
git clone https://github.com/jacksonfdam/medusa-nexus.git
cd medusa-nexus
scripts/setup.sh
 
# 3. Wire env vars into your shell rc (the installer prints the line).
echo 'source ~/.mnexus/env.sh' >> ~/.zshrc
source ~/.mnexus/env.sh
 
# 4. Verify.
mnexus doctor

What `mnexus doctor` checks

The doctor walks every registered engine and prints a status row per row:


🔱 engine             status   version   note
────────────────────────────────────────────────────────────────────
   adb                ● OK     34.0.5    /usr/local/bin/adb
   jadx               ● OK     1.5.0     /Users/.../tools/jadx
   apktool            ● OK     2.9.3     /Users/.../tools/apktool
   ghidra             ● OK     11.1.2    /Users/.../tools/ghidra_11.1.2_PUBLIC
   mobsf              ● OK     4.0.6     http://localhost:8000
   frida              ● OK     16.4.10
   burp               ● MISS             MNEXUS_BURP_API_KEY not set
   moxy               ● OK     latest    http://localhost:5000
   playintel          ● OK     —         ready
   vphone             ● MISS             super-tart-vphone not configured

MISS is not a failure. The pipeline runs every engine that’s available and routes around the ones that aren’t. Run mnexus doctor --env to see which MNEXUS_* variables actually reached the process (the most common cause of an unexpected MISS).

iOS extras


scripts/setup.sh --ios-tools

Installs bagbak, ldid, and clones frida-ios-dump. See the iOS workflow → for what each one does.

Burp / Caido / Moxy

You only need one. The lightest path is Moxy:


scripts/setup.sh --moxy

This starts the Moxy Docker container, waits for the dashboard to come up, extracts the mitmproxy CA, pushes it to your connected ADB device as a system trust anchor (root required), and writes MNEXUS_MOXY_* + MNEXUS_PROXY_FLAVOR=moxy to ~/.mnexus/env.sh. See Moxy →.

For Burp, install Burp Suite Pro manually, then run scripts/setup.sh to drop the REST API extension into your Burp configuration directory.

For Caido, install Caido yourself, generate an API token at Workbench → Settings → Tokens, then add it to env.sh:


export MNEXUS_PROXY_FLAVOR=caido
export MNEXUS_CAIDO_URL=http://localhost:8080
export MNEXUS_CAIDO_API_KEY=cai_…

Updating


cd medusa-nexus
git pull
scripts/setup.sh        # idempotent — only re-installs what changed

The installer skips engines that are already at the right version. Ghidra and the JADX release tarball are pinned to specific versions in the script header — bumping the floor requires a manual edit.

Uninstall


rm -rf ~/.mnexus
# then remove the `source ~/.mnexus/env.sh` line from your shell rc

The repo checkout itself can be deleted with a plain rm -rf — nothing in MedusaNexus writes outside of $MNEXUS_HOME.

Next: Environment variables → to understand what every MNEXUS_* knob does.