NovaCoder AI
NovaCoder AI is an open-source AI coding assistant for Visual Studio Code — built for developers who want full control over their AI models, data, and workflow.
Use any AI provider (OpenAI, Anthropic, Claude, Gemini, OpenRouter, DeepSeek, Ollama, LM Studio) with your own API key. Chat, explain, refactor, fix errors, generate tests, and apply code changes directly to your editor — without leaving VS Code.
Features
AI Chat Panel
- Streaming chat responses with syntax-highlighted code blocks
- Apply button on every code block — inserts or replaces code in your active file
- New File button — opens generated code in a new editor tab
- Copy button on every code block and message
- Typing indicator and streaming cursor while AI responds
- Chat history with search, pin, rename, and delete
- Explain Code — understand what any code does
- Fix Errors — diagnose and patch problems
- Refactor — clean up and improve code quality
- Generate Tests — write unit tests automatically
- Add Docs — insert documentation comments
- Optimize — find and fix performance bottlenecks
Right-Click Commands
All commands available via the Command Palette (Ctrl+Shift+P) and right-click context menu:
| Command |
Description |
| NovaCoder: Open AI Chat |
Open the sidebar chat |
| NovaCoder: Explain Code |
Explain selected code |
| NovaCoder: Refactor Selection |
Refactor selected code |
| NovaCoder: Fix Errors |
Fix errors in the file |
| NovaCoder: Generate Tests |
Write tests for selected code |
| NovaCoder: Create Documentation |
Add docs to selected code |
| NovaCoder: Optimize Performance |
Suggest performance improvements |
| NovaCoder: Create Commit Message |
Generate a git commit message |
| NovaCoder: Ask AI About File |
Chat about the current file |
Multi-Provider Support
| Provider |
Models |
| OpenAI |
GPT-4o, GPT-4.1, GPT-4 Turbo, o1, o3-mini, and more |
| Anthropic |
Claude Opus 4, Claude Sonnet 4, Claude Haiku 4.5 |
| Google Gemini |
Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.0 Flash |
| OpenRouter |
200+ models including free tiers (Nemotron, GPT-OSS) |
| DeepSeek |
DeepSeek Chat, DeepSeek Reasoner |
| Ollama |
Llama 3.3, Mistral, Phi-4, Gemma 3, Qwen 2.5, DeepSeek R1 |
| LM Studio |
Any local model running in LM Studio |
Secure by Design
- API keys stored in VS Code SecretStorage — never written to disk or config files
- Local-first architecture — workspace index and memory stay on your machine
- Terminal safety policy — destructive commands are blocked and require explicit approval
Autonomous Agent
The Agent tab runs a plan-execute-validate loop:
- Breaks your task into steps
- Executes tools (file read/write, search, terminal, git diff)
- Validates output and retries on failure
- Returns a structured summary
Workspace Awareness
NovaCoder automatically collects context from:
- Active file and current selection
- Open editor tabs
- Diagnostics (errors and warnings)
- Git diff
- Semantic workspace index
Installation
- Install from the VS Code Marketplace or download the
.vsix file
- Open the NovaCoder icon in the Activity Bar (left sidebar)
- Go to Settings tab and enter your API key for your preferred provider
- Click Save Settings and start chatting
Configuration
All settings are available under Settings → NovaCoder AI or in settings.json:
| Setting |
Default |
Description |
novacoder.defaultProvider |
openai |
Default AI provider |
novacoder.defaultModel |
gpt-4.1-mini |
Default model ID |
novacoder.temperature |
0.2 |
Sampling temperature (0–2) |
novacoder.maxTokens |
2000 |
Max response tokens |
novacoder.enableAutocomplete |
true |
Enable inline AI completions |
novacoder.contextTokenBudget |
12000 |
Max context tokens per request |
novacoder.indexing.enabled |
true |
Enable semantic workspace indexing |
novacoder.terminal.requireApproval |
true |
Require approval before running terminal commands |
Provider Setup
OpenAI
- Get your key at platform.openai.com
- In Settings tab → select OpenAI → enter API key → Save
Anthropic (Claude)
- Get your key at console.anthropic.com
- In Settings tab → select Anthropic → enter API key → Save
Google Gemini
- Get your key at aistudio.google.com
- In Settings tab → select Google Gemini → enter API key → Save
OpenRouter (200+ models, free tiers available)
- Get your key at openrouter.ai
- In Settings tab → select OpenRouter → enter API key → Save
Ollama (local, free, no key needed)
- Install Ollama from ollama.com
- Run
ollama pull llama3.3 in terminal
- In Settings tab → select Ollama → Save (no API key required)
LM Studio (local, free, no key needed)
- Install LM Studio from lmstudio.ai
- Download a model and start the local server
- In Settings tab → select LM Studio → Save (no API key required)
Using Code Actions
When NovaCoder responds with a code block, you get three action buttons:
- Apply — Inserts the code at your cursor, or replaces your current selection
- New File — Opens the code in a new editor tab beside the current file
- Copy — Copies the code to clipboard
Keyboard Shortcuts
| Action |
Shortcut |
| Send message |
Ctrl+Enter |
Building from Source
git clone https://github.com/novacoder/novacoder-ai
cd novacoder-ai
npm install
npm run build
To run in development:
npm run dev
# Then press F5 in VS Code to launch the Extension Development Host
To create a .vsix package:
npm run package
Privacy
- No telemetry, no analytics, no data collection
- API calls go directly from your machine to the provider you configure
- Chat history and workspace index are stored locally in VS Code's extension storage
- API keys are stored in VS Code SecretStorage (OS keychain-backed)
License
MIT — free for personal and commercial use.
Contributing
Issues and pull requests welcome at github.com/novacoder/novacoder-ai.
