A clear view of your code.
Inspect selected project files, source hashes, and the actual investigation scope in the local workspace.
Meet the Project Explorer →YOUR CODE. A SECOND PERSPECTIVE.Understand deeply.
Understand deeply.
Change confidently.
Meet DevTwin, your engineering investigation workspace. Connect a project, follow the evidence, and turn a question into a change you can review.
01 / Inspect your project02 / Explore 13 workflows03 / Review & verify
engineering / workspace DEVTWIN
project.connect()question → evidenceinvestigation.prepare({ workflows: 13, source: your selection, changes: review first})
↳
Make the next step inspectable.Source references · reviewed patches · isolated checks
Thirteen perspectives.
Trace features, diagnose bugs, and review security, compatibility, performance, and delivery evidence.
Explore all workflows →Review before confidence.
Review proposed patches and separately approve original/candidate Python tests in Docker.
See the local setup →
THE PERSON BEHIND DEVTWINBUILT BY A DEVELOPER, FOR DEVELOPERS
Hi, I’m Raimal Raja.
Machine Learning Engineer & creator of DevTwin.
I’m a final-year Computer Science student at the University of Sindh, Laar Campus, Badin, with an expected graduation in 2027. My work spans machine learning, LLM fine-tuning, computer vision and Python backend applications.
I work remotely as a Data Administrator at KV-Tech and previously completed a paid Python Developer internship at SignDevOps. I build with Python, FastAPI, TensorFlow and PyTorch, bringing data pipelines and inference applications together.
Machine learningPython & APIsApplied AIOracle certified
Biography based on the supplied CV and public GitHub profile. Photo: GitHub profile.ONE PACKAGE. YOUR AGENT.
Make it part of your workflow.
Choose your runtime and installation format.
Download plugin package ↓
Local manifests and package integrity are checked.
Codex loading checked; Claude execution pending.
SETUP GUIDE CODEX
Get the package
Install in your runtime
TERMINALStart with a real task
Get to know
your code.
A precise source selection. A clear question. An investigation you can inspect.
This public site introduces the workspace. GitHub Pages does not run the Python backend or Docker worker. Project uploads, GitHub imports, model calls, patch proposals, and isolated checks are available in the separately configured local app.
GitHub or project ZIP
Import selected UTF-8 files. GitHub imports are read-only and pinned to a commit. Limits: 30 files, 16,000 selected bytes, and a 2 MiB ZIP.
Know what is included
Preview literal source, file sizes, hashes, omitted files, and Python test prerequisites before provider submission.
Your review comes first
Provider submission, patch export, test execution, and the bounded repair loop have explicit review and approval controls.
Local execution. Public setup guide.
Clone the source and prepare Docker with Linux containers and Python 3.10+. The OpenRouter helper prompts privately for your provider key. Review its data policy before submitting source.
git clone https://github.com/Raimal-Raja/devtwin-engineer.git cd devtwin-engineer docker pull python:3.12-slim docker pull docker:cli python scripts/prepare_docker.py . python scripts/start_openrouter.py
After startup, open http://127.0.0.1:8765/ on your computer and enter your workspace session token from .env.docker. Never enter a provider API key in that browser field.
Ask better questions.
Follow the evidence.
Explore the 13 workflow guides and prepare an agent task. Execute source investigations and reviewed repairs in your local workspace.
CHOOSE YOUR STARTING POINT
Workflow catalog 13
No matching workflows. Try another search or category.
READY FOR YOUR NEXT INVESTIGATION
A clear task. A useful handoff.
Prepare a prompt for the runtime you selected. Your task text stays in this page.
AGENT PROMPT