Jos: Code Assist
AI-powered coding assistant for VS Code with free AI models via OpenRouter, Groq, and Z.AI. No Python required.
Features
- Multiple Free Providers - OpenRouter, Groq, and Z.AI, each with their own model lineup
- Agent Mode - Let the assistant read, search, and edit files across your workspace with your approval (Z.AI GLM-4.7/4.5 Flash models)
- Clean Chat Interface - Side panel chat with full markdown support
- Real Streaming - Responses stream word-by-word via direct API calls
- Session Management - Multiple conversations, export to Markdown
- Secure Storage - API key stored securely in VS Code
- Fast Switching - Change platforms and models on the fly
- History Persistence - All chats saved locally
- No Dependencies - Pure TypeScript, no Python or pip needed
Requirements
- VS Code 1.75.0 or higher
- A free API key from at least one supported platform:
Quick Start
- Install the extension
- Click the Jos: Code Assist icon in the left sidebar
- Choose a platform (OpenRouter, Groq, or Z.AI) and enter its API key when prompted
- Start chatting immediately
Agent Mode
Toggle Agent on in the chat panel to let the assistant work directly with your codebase instead of just answering in chat. It can:
- List and search files across the workspace
- Read files (with pagination for large files)
- Check current compiler/linter errors
- Propose file edits or create new files
Every change is shown as a diff before anything is written, requires your confirmation, and can be undone (single file or all at once) from the chat panel. Agent Mode is currently only available on Z.AI's GLM-4.7 Flash and GLM-4.5 Flash models — switching to a different platform or an incompatible model disables it automatically.
Commands
| Command |
Shortcut |
| Open Chat |
Ctrl+Shift+J / Cmd+Shift+J |
| Configure API Key |
- |
| New Chat Session |
- |
| Clear Current Chat |
- |
| Export Chat to Markdown |
- |
| Reset Extension |
- |
⚠️ Privacy Notice
These models are provided for testing and demo purposes. Please avoid sharing any personal or confidential information in your conversations, including passwords, API keys, private code, or sensitive business data.
Available Models
OpenRouter (14 verified working)
Google (1 model)
| Model |
Context Window |
Input Modalities |
Output Modalities |
| Gemma 4 26B A4B IT |
256K tokens |
Text, Image, Video |
Text |
NVIDIA (6 models)
| Model |
Context Window |
Input Modalities |
Output Modalities |
| Nemotron 3 Super 120B |
1M tokens |
Text |
Text |
| Nemotron 3 Nano 30B |
256K tokens |
Text |
Text |
| Nemotron 3 Nano Omni 30B Reasoning |
256K tokens |
Text, Audio, Image, Video |
Text |
| Nemotron 3.5 Content Safety |
128K tokens |
Text, Image |
Text |
| Nemotron Nano 12B V2 VL |
128K tokens |
Text, Image |
Text |
| Nemotron Nano 9B V2 |
128K tokens |
Text |
Text |
⚠️ NVIDIA Data Notice: Please do not upload any confidential information. You may upload images, which NVIDIA and its service providers will use solely for providing this demo experience. Your use is logged for security purposes and anonymous session data may be used to improve NVIDIA products and services, including network performance metrics, AI-generated outputs, and audio transcriptions. The logged session data for improvement purposes is not linked to your identity or any persistent identifier. For more information, see NVIDIA's Privacy Policy. By using these models, you consent to NVIDIA's collection, recording, and use of such information and the NVIDIA API Trial Terms of Service.
OpenAI (1 model)
| Model |
Context Window |
Input Modalities |
Output Modalities |
| GPT OSS 20B |
131K tokens |
Text |
Text |
Poolside (2 models) — Default: Laguna M.1
| Model |
Context Window |
Input Modalities |
Output Modalities |
| Laguna M.1 ⭐ |
262K tokens |
Text |
Text |
| Laguna XS.2 |
262K tokens |
Text |
Text |
Other Providers (4 models)
| Model |
Provider |
Context Window |
Input Modalities |
Output Modalities |
| North Mini Code |
Cohere |
256K tokens |
Text |
Text |
| Owl Alpha |
OpenRouter |
1M tokens |
Text |
Text |
| LFM 2.5 1.2B Thinking |
LiquidAI |
33K tokens |
Text |
Text |
| LFM 2.5 1.2B Instruct |
LiquidAI |
33K tokens |
Text |
Text |
Groq
Fast inference for Llama 3.3/3.1 and GPT-OSS models.
Z.AI
GLM-4.7 Flash and GLM-4.5 Flash, plus Qwen3.6 — required if you want to use Agent Mode.
Architecture
TypeScript Extension → OpenRouter / Groq / Z.AI APIs (direct HTTPS)
No Python bridge, no subprocess spawning, no pip dependencies.
Troubleshooting
❌ "Invalid API key"
- OpenRouter keys start with
sk-or- — get a new one at openrouter.ai/keys
- Groq and Z.AI keys follow their own formats — get them from console.groq.com/keys or z.ai
- Use command: Jos: Code Assist: Configure API Key to switch platforms or update a key
❌ Agent Mode toggle is disabled or grayed out
- Agent Mode only works with Z.AI's GLM-4.7 Flash or GLM-4.5 Flash models
- Switch platform to Z.AI and select one of those models via Jos: Code Assist: Configure API Key
❌ Extension icon not visible
- Press
Ctrl+Shift+P (Cmd+Shift+P on Mac)
- Type Jos: Code Assist: Open Chat
- Press Enter
❌ No response from a model
- Switch to a different model using the dropdown
- Check your internet connection
- Verify your API key is valid at openrouter.ai/keys
Still having issues?
- Open VS Code Developer Tools:
Help → Toggle Developer Tools
- Check the Console tab for error messages
System Requirements
| Requirement |
Version |
| VS Code |
1.75.0+ |
| Internet |
Required |
Uninstallation
- Go to Extensions (
Ctrl+Shift+X)
- Find Jos: Code Assist and click Uninstall
- Optionally run Jos: Code Assist: Reset Extension to clear API key and chat history
License
See LICENSE file for details.
Credits
Built with OpenRouter's free AI models.
Powered by JOSENTITY
For support, contact us at support@josentity.com
