Everything Gencode NG does
A complete AI coding surface for VS Code — four modes, deep codebase understanding, and complete model freedom.
Autocomplete Mode
Ghost-text code suggestions appear inline as you type, mirroring the fluid experience of GitHub Copilot — but powered by whichever model you choose.
Chat Mode
A dedicated sidebar panel for engineering questions, file explanations and debugging conversations that stay in your editor.
Edit Mode
Highlight any code block and issue natural language instructions to modify, optimize or refactor it inline with a reviewable diff.
Agent Mode
Autonomous multi-step tasks: reading files, executing terminal commands and managing complex edits across your codebase.
Key workflow features
Codebase Context (RAG)
Reference your entire codebase, specific files or docs in chat using @ commands. Gencode NG indexes locally so the model understands your architecture.
Inline Diffs
Every generated or refactored change renders as a side-by-side or inline diff. You audit and accept — nothing lands silently.
Slash Commands
Built-in shortcuts like /edit, /comment and /fix trigger pre-defined actions instantly without typing long descriptions.
Terminal Error Tracking
Terminal output and compiler errors are captured automatically, offering one-click troubleshooting the moment something fails.
Bring your own intelligence
Vendor Agnostic
Plug in Anthropic Claude, OpenAI GPT, Google Gemini or xAI — swap providers whenever you want.
Local & Offline
Pair with Ollama to run DeepSeek-R1, Qwen-2.5-Coder and other open models entirely offline. Proprietary code never leaves your machine.
Multi-Model Splitting
Assign different models to different tasks — a fast model for autocomplete and a heavy reasoner for chat, in parallel.
Configuration Control
Prompts, endpoints and settings live in a transparent config.json or YAML file that you own and version.
TokenSaver
Advanced prompt optimization, context and history management, and output formatting controls drastically cut token usage.
