diff --git a/CLAUDE.md b/CLAUDE.md
index 3dd0afd..8586cff 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -13,7 +13,7 @@ The goal is that `docker compose up` gives you a working setup where Claude Desk
 
 ## Status
 
-The repo is at an early stage (no code yet). The layout below is the **intended** structure. Update this file when the real structure differs.
+Early stage. What exists: a single image (`Dockerfile`, `launch-apps.sh`) with Claude Desktop and Ghidra, started by `run.sh`; no compose setup yet. The layout below is the **intended** structure. Update this file when the real structure differs.
 
 ## Intended layout
 
@@ -37,7 +37,7 @@ The repo is at an early stage (no code yet). The layout below is the **intended*
 
 ## Architecture
 
-- **One service per container.** Claude Desktop, Ghidra and the RE MCP server each get their own image, connected on one compose network.
+- **Claude Desktop and Ghidra share one container** (the root `Dockerfile`, started by `run.sh`), at the user's request. Ghidra is the official GitHub release (version, build date and SHA-256 are build args), runs on Ubuntu's `openjdk-25-jdk`, lives in `/opt/ghidra` and starts together with Claude Desktop. Further services such as the RE MCP server may still get their own containers.
 - **Shared workspace volume.** Every container mounts the workspace at `/workspace`, so a file path means the same thing to Claude, Ghidra and the RE MCP. `run.sh` asks for the host folder on each start in the desktop's folder picker (`zenity`, or `kdialog` on KDE; terminal prompt if neither exists), opening at the last folder used, else `./workspace`. Skip it with `--workspace DIR` or `WORKSPACE=DIR`. It refuses `/` and `$HOME`. The image symlinks `~/workspace` to `/workspace` for Claude Desktop's file picker.
 - **GUI forwarding.** Claude Desktop and Ghidra are GUI apps. Default to X11 forwarding (`/tmp/.X11-unix` + `DISPLAY`). Support Wayland through XWayland. Keep a noVNC/VNC fallback in mind for headless hosts. Never solve display access with a blanket `xhost +`. Scope it (`xhost +local:docker` or `+SI:localuser:...`).
 - **GhidraMCP wiring.** GhidraMCP has two parts:
@@ -45,7 +45,7 @@ The repo is at an early stage (no code yet). The layout below is the **intended*
   - a Python MCP bridge (`bridge_mcp_ghidra.py`) that speaks MCP to Claude and forwards calls to that HTTP server.
 
   The bridge is configured with the Ghidra server URL (e.g. `http://ghidra:8080/`). The plugin must be enabled in the CodeBrowser and a program must be open, or the bridge has nothing to talk to.
-- **Container lifetime follows the window.** Closing Claude Desktop's window only hides the app (GNOME shows no tray icon to bring it back). The image's entrypoint `launch-claude.sh` watches the app's X windows: when none is open or minimized for ~3 s, it quits the app and the container exits. `run.sh` uses `--init` so signals and zombie processes are handled.
+- **Container lifetime follows the apps.** The entrypoint `launch-apps.sh` starts Claude Desktop and Ghidra and stops both, ending the container, when either is closed. Ghidra counts as closed when its process exits (closing the project window quits it; CodeBrowser or Help windows don't). Claude Desktop's window only hides the app on close (GNOME shows no tray icon), so its X windows are watched: none open or minimized for ~3 s counts as closed. `run.sh` uses `--init` so signals and zombie processes are handled.
 - **MCP transport between containers.** Claude Desktop starts local MCP servers as stdio subprocesses. Because the servers live in other containers, pick one approach and apply it consistently:
   - *Preferred:* run the MCP servers with an HTTP/SSE transport in their own containers. Bridge them into Claude Desktop's stdio config with a small proxy (e.g. `mcp-remote` or `mcp-proxy`) installed in the Claude Desktop image.
   - *Alternative:* install the bridges inside the Claude Desktop image and point them at the other containers over the network. This works well for the GhidraMCP bridge, which is just Python + HTTP.
diff --git a/Dockerfile b/Dockerfile
index 07158b3..068a4e1 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -33,6 +33,25 @@ RUN apt-get update \
         fonts-dejavu-core \
  && rm -rf /var/lib/apt/lists/*
 
+# Ghidra from the official GitHub release, checked against the SHA-256 in the
+# release notes: https://github.com/NationalSecurityAgency/ghidra/releases
+# Ghidra 12.2 needs JDK 25 (application.java.min=25). The full JDK, not the
+# headless one, because Ghidra's GUI needs the AWT/X11 libraries.
+ARG GHIDRA_VERSION=12.2
+ARG GHIDRA_BUILD_DATE=20261016
+ARG GHIDRA_SHA256=2c2cd776db3c1f6a097ba418922378ac3649cadd5e844a6f5f088ae628a7be67
+RUN apt-get update \
+ && apt-get install -y --no-install-recommends openjdk-25-jdk unzip \
+ && rm -rf /var/lib/apt/lists/* \
+ && curl -fsSLo /tmp/ghidra.zip \
+        "https://github.com/NationalSecurityAgency/ghidra/releases/download/Ghidra_${GHIDRA_VERSION}_build/ghidra_${GHIDRA_VERSION}_PUBLIC_${GHIDRA_BUILD_DATE}.zip" \
+ && echo "${GHIDRA_SHA256}  /tmp/ghidra.zip" | sha256sum -c - \
+ && unzip -q /tmp/ghidra.zip -d /opt \
+ && rm /tmp/ghidra.zip \
+ && ln -s "/opt/ghidra_${GHIDRA_VERSION}_PUBLIC" /opt/ghidra \
+ && ln -s /opt/ghidra/ghidraRun /usr/local/bin/ghidra
+ENV GHIDRA_INSTALL_DIR=/opt/ghidra
+
 # Ubuntu 24.04 ships a default "ubuntu" user on UID 1000; replace it.
 RUN if id ubuntu >/dev/null 2>&1; then userdel -r ubuntu; fi \
  && groupadd -g "${USER_GID}" "${USER_NAME}" \
@@ -40,16 +59,17 @@ RUN if id ubuntu >/dev/null 2>&1; then userdel -r ubuntu; fi \
  && install -d -o "${USER_UID}" -g "${USER_GID}" /workspace \
  && ln -s /workspace "/home/${USER_NAME}/workspace"
 
-COPY --chmod=755 launch-claude.sh /usr/local/bin/launch-claude
+COPY --chmod=755 launch-apps.sh /usr/local/bin/launch-apps
 
 USER ${USER_NAME}
 WORKDIR /home/${USER_NAME}
 
-# launch-claude quits the app (and so the container) when its window closes.
+# launch-apps starts Claude Desktop and Ghidra and stops both (and so the
+# container) when either one is closed.
 # No keyring: tell Electron not to look for one. Session data lives only in
 # the container's filesystem, which is discarded on exit (run with --rm), so
 # the user signs in on every start.
 # --no-sandbox: Docker's default seccomp profile blocks the namespaces
 # Chromium's sandbox needs, so the zygote dies on start. The container's own
 # seccomp/namespaces stay in place as the boundary to the host instead.
-CMD ["launch-claude", "--password-store=basic", "--no-sandbox"]
+CMD ["launch-apps", "--password-store=basic", "--no-sandbox"]
diff --git a/launch-apps.sh b/launch-apps.sh
new file mode 100755
index 0000000..279c6d5
--- /dev/null
+++ b/launch-apps.sh
@@ -0,0 +1,68 @@
+#!/usr/bin/env bash
+# Container entrypoint: start Claude Desktop and Ghidra, and stop both (so the
+# container exits) as soon as either one is closed.
+#
+# - Ghidra exits when its project window is closed, so its process ending is
+#   the signal. Closing other Ghidra windows (CodeBrowser, Help) doesn't count.
+# - Closing Claude Desktop's window only hides the app (it keeps running for
+#   the tray icon, which GNOME doesn't show). So its X windows are watched and
+#   it counts as closed once none of them is open or minimized.
+#
+# Arguments are passed to Claude Desktop.
+
+set -u
+# Job control puts each app in its own process group, so stopping an app also
+# stops everything it started (Ghidra's launch script runs java as a child).
+set -m
+
+claude-desktop "$@" &
+claude=$!
+"${GHIDRA_INSTALL_DIR:-/opt/ghidra}/support/launch.sh" fg jdk Ghidra "" "" ghidra.GhidraRun &
+ghidra=$!
+
+stop_all() {
+    trap - TERM INT HUP
+    echo "launch-apps: $1, stopping" >&2
+    # Claude's main process shuts down its own helpers; killing its whole
+    # group would take out the GPU process first and trigger restart loops.
+    # Ghidra's group is needed to reach java behind the launch script.
+    kill -TERM "$claude" 2>/dev/null
+    kill -TERM -- "-$ghidra" 2>/dev/null
+    # Give both a few seconds to shut down cleanly before the container ends.
+    for _ in {1..10}; do
+        kill -0 "$claude" 2>/dev/null || kill -0 -- "-$ghidra" 2>/dev/null || break
+        sleep 0.5
+    done
+    exit 0
+}
+trap 'stop_all "received stop signal"' TERM INT HUP
+
+host="$(hostname)"
+
+# True if any of this container's Claude windows is Normal (shown) or Iconic
+# (minimized). A closed window is Withdrawn or loses WM_STATE altogether.
+# WM_CLIENT_MACHINE filters out Claude windows from other containers or apps
+# on the same X server.
+claude_window_open() {
+    local w props
+    for w in $(xwininfo -root -tree 2>/dev/null \
+               | awk '/\("com\.anthropic\.claude" "com\.anthropic\.Claude"\)/ {print $1}'); do
+        props="$(xprop -id "$w" WM_CLIENT_MACHINE WM_STATE 2>/dev/null)" || continue
+        [[ "$props" == *"WM_CLIENT_MACHINE(STRING) = \"$host\""* ]] || continue
+        [[ "$props" =~ "window state: "(Normal|Iconic) ]] && return 0
+    done
+    return 1
+}
+
+seen=0    # only arm the window check after Claude's window has appeared
+misses=0  # require a few consecutive misses to ride out brief remaps
+while :; do
+    sleep 1 & wait $!
+    kill -0 "$ghidra" 2>/dev/null || stop_all "Ghidra exited"
+    kill -0 "$claude" 2>/dev/null || stop_all "Claude Desktop exited"
+    if claude_window_open; then
+        seen=1 misses=0
+    elif (( seen )) && (( ++misses >= 3 )); then
+        stop_all "Claude Desktop window closed"
+    fi
+done
diff --git a/launch-claude.sh b/launch-claude.sh
deleted file mode 100755
index af8001e..0000000
--- a/launch-claude.sh
+++ /dev/null
@@ -1,46 +0,0 @@
-#!/usr/bin/env bash
-# Container entrypoint: run Claude Desktop and stop when its window is closed.
-#
-# Closing the window only hides the app (it keeps running for the tray icon,
-# which GNOME doesn't show), so the container would keep running invisibly.
-# This watches the app's X windows and quits the app once none of them is
-# open or minimized. The container then exits with it.
-
-set -u
-
-claude-desktop "$@" &
-app=$!
-
-stop_app() { kill -TERM "$app" 2>/dev/null; wait "$app"; exit $?; }
-trap stop_app TERM INT HUP
-
-host="$(hostname)"
-
-# True if any of this container's Claude windows is Normal (shown) or Iconic
-# (minimized). A closed window is Withdrawn or loses WM_STATE altogether.
-# WM_CLIENT_MACHINE filters out Claude windows from other containers or apps
-# on the same X server.
-window_open() {
-    local w props
-    for w in $(xwininfo -root -tree 2>/dev/null \
-               | awk '/\("com\.anthropic\.claude" "com\.anthropic\.Claude"\)/ {print $1}'); do
-        props="$(xprop -id "$w" WM_CLIENT_MACHINE WM_STATE 2>/dev/null)" || continue
-        [[ "$props" == *"WM_CLIENT_MACHINE(STRING) = \"$host\""* ]] || continue
-        [[ "$props" =~ "window state: "(Normal|Iconic) ]] && return 0
-    done
-    return 1
-}
-
-seen=0    # only arm the check after the first window has appeared
-misses=0  # require a few consecutive misses to ride out brief remaps
-while kill -0 "$app" 2>/dev/null; do
-    sleep 1 & wait $!
-    if window_open; then
-        seen=1 misses=0
-    elif (( seen )) && (( ++misses >= 3 )); then
-        echo "launch-claude: window closed, quitting Claude Desktop" >&2
-        stop_app
-    fi
-done
-
-wait "$app"
diff --git a/run.sh b/run.sh
index 333d493..2aab484 100755
--- a/run.sh
+++ b/run.sh
@@ -192,7 +192,7 @@ printf '%s\n' "$WORKSPACE" > "$last_workspace_file"
 # is discarded and Claude asks for a login on every start.
 # No :z on the workspace mount: label=disable already lets the container use
 # it, and relabeling an arbitrary host folder could break other apps.
-info "Starting Claude Desktop in container '$CONTAINER' (workspace: $WORKSPACE -> /workspace)"
+info "Starting Claude Desktop and Ghidra in container '$CONTAINER' (workspace: $WORKSPACE -> /workspace)"
 "$RUNTIME" run -d "${run_args[@]}" \
     --name "$CONTAINER" \
     -v "$WORKSPACE:/workspace" \