ghidrasql-ai

In a world where large corporations can simply steal all your code or data to train AI and sell it back to you, it's time hacking and hackers regained their political conscience. Hackers have grown politically inept in the pursuit of profit while our tools have grown better. This is a docker build of Elias's wonderful projects ghidrasql and libghidra. Reverse engineering is hard and slow to do, any effort to make it that little bit easier should be celebrated. Leverage the tools of oppression and surveillance against your adversary (big tech) - in this case use AI against them. All information deserves to be free so set your AI minions on their proprietary software and hack the planet. They can run but they can't hide from the ghidra.

Docker release for running ghidrasql over a live LibGhidraHost / Ghidra stack.

Two long-running containers expose:

• LibGhidraHost RPC at http://127.0.0.1:18080
• SQL HTTP API at http://127.0.0.1:8081/query

Prerequisites

• Docker and Docker Compose
• ~8 GB RAM allocated to Docker (build and analysis are memory-heavy)
• ~4 GB free disk for image build
• On Windows: Docker Desktop with WSL2 backend (recommended)

Pinned versions

Component	Version / ref
Ghidra	12.0.4 PUBLIC (20260303) — validated in original ghidra-ai stack
libghidra	main
libxsql	ea11622eeec5ac7e5988364ebfaffefccb1bb3f4
ghidrasql	main
ghidrasql-skills	main

Ghidra 12.0.4 matches the working dev image (ghidra-base:local). libghidra requires 12.0.4+; ghidrasql upstream docs mention 12.1+ but this release was validated on 12.0.4. See docs/BUILD.md for version notes and the historical MemoryBlock patch errors.

Quick start

1. Build images

Linux / macOS / Git Bash: This does everything needed before running docker compose up -d

./build.sh

Windows (cmd or PowerShell):

build.bat

This builds ghidrasql-ai-base:latest (Ghidra 12.0.4), ghidrasql-ai:latest, and compiles samples/hello when gcc is available.

Use a local extracted Ghidra tree instead of downloading:

./build.sh --local ../ghidra_12.0.4_PUBLIC

build.bat --local ..\ghidra_12.0.4_PUBLIC

2. Import a demo binary

Compile the sample (Linux/macOS):

gcc -o samples/hello samples/hello.c

Windows (MSVC):

cl samples\hello.c /Fe:samples\hello.exe
copy samples\hello.exe samples\hello

Or with MinGW:

gcc -o samples\hello samples\hello.c

Import and start (Linux/macOS):

cp .env.example .env
./scripts/import-demo.sh
# or manually: docker compose down && docker compose --profile import run --rm ghidrasql-import && docker compose up -d

If compose fails with dependency ghidrasql-libghidra failed to start, run ./scripts/diagnose.sh.

Windows:

scripts\win\import-demo.bat

3. Smoke test

curl -sS -X POST http://127.0.0.1:8081/query \
  --data "SELECT COUNT(*) AS funcs FROM funcs"

Windows:

scripts\win\smoke-test.bat

Daily operations

Start services:

cp .env.example .env   # first time only
docker compose up -d

Stop services:

docker compose down

Configure the active Ghidra project/program in .env:

GHIDRA_PROJECT_NAME=hello_demo
GHIDRA_PROGRAM_NAME=hello

Documentation

Doc	Purpose
docs/BUILD.md	Build troubleshooting — version choice, common errors
docs/OPERATIONS.md	Quick reference — import, SQL examples, troubleshooting
docs/ghidrasql_operational_guide_v1.md	Full operational guide — schema details, workflows, diagnostics
docs/AI.md	Using with AI agents — Cursor, Claude Code, ghidrasql-skills, SQL contract

Using with AI

This stack exposes a SQL HTTP API that coding agents query instead of opening Ghidra directly:

curl -sS -X POST http://127.0.0.1:8081/query \
  --data "SELECT name, printf('0x%X', address) AS addr FROM funcs LIMIT 10"

Install ghidrasql-skills into your agent (bundled in the image at /opt/ghidrasql-skills) and follow docs/AI.md for Cursor, Claude Code, and Codex setup.

See docs/OPERATIONS.md for day-to-day Docker operations.

Windows scripts

Script	Purpose
build.bat	Build both Docker images
scripts\win\up.bat	Create .env if missing; start services
scripts\win\down.bat	Stop services
scripts\win\import-demo.bat	Import samples\hello and start services
scripts\win\smoke-test.bat	Run a baseline SQL query

Architecture

ghidrasql-libghidra   Ghidra headless + LibGhidraHost  :18080
ghidrasql-server      ghidrasql SQL proxy              :8081

ghidrasql-server connects to ghidrasql-libghidra by URL and does not open the Ghidra project itself.

Native Windows (without Docker)

This repository packages the Docker workflow only. For a native Windows install (Ghidra zip + VS 2022 + CMake + Gradle), follow the upstream runbooks:

• ghidrasql install-prompt.md
• libghidra README

License

Components retain their upstream licenses (Ghidra: Apache 2.0; ghidrasql/libghidra: MPL 2.0).