Project-aware AI Chat
Ask about the current codebase and get answers with relevant source files.
CodePilot AI
Local AI Coding Assistant
Built for developers. Powered locally.
CodePilot AI v1.0
VS Code extension · Local AI · RAG
Debug, understand and improve codewithout leaving VS Code.
CodePilot AI combines local LLMs, project-aware RAG, VS Code diagnostics and safe fix previews into one coding assistant built directly inside your editor.
✓ Local Ollama models✓ LangChain RAG✓ Safe fix previews
CODEPILOT AI
script.js
function greet(name) {
if (name) {
console.log("Hello " + name);
}
}
Project-aware answer
The greet function in script.js checks whether a name exists and prints a greeting to the console.
See it in action
Real CodePilot workflows inside VS Code.
Project-aware answers, source navigation and safe fix previews are all available directly from the editor.
Error Detection & Safe Fixes
Review generated fixes before applying changes to your source code.
Tip: CodePilot keeps AI-generated code visible before applying it, so you stay in control of every change.
Core features
More context. Safer fixes. Less switching.
01
Project-Aware RAG
Ask questions about your codebase and get answers grounded in relevant project files.
02
Safe AI Fixes
Preview generated fixes before applying them, with safeguards against stale or unsafe edits.
03
Local by Default
Runs with Ollama and local models, keeping the workflow independent of paid cloud APIs.
How it works
Context in. Useful assistance out.
CodePilot combines editor context, project retrieval and local inference before showing the result inside VS Code.
1
VS Code detects context and diagnostics
2
CodePilot retrieves relevant project files
3
Local Ollama model generates an answer or fix
4
You review the response before applying changes
Privacy-first workflow
Your code stays in your local development loop.
CodePilot is designed around local Ollama inference. Project context, diagnostics and code snippets are sent to the locally running model instead of a paid cloud AI API in the current setup.
Local inference
Qwen runs through Ollama on your machine.
Local embeddings
Project retrieval uses local embeddings for RAG.
Architecture
Context from VS Code, retrieval from your project, answers from a local model.
1
VS Code Context
Active file, cursor, function and diagnostics
2
RAG Retrieval
Relevant project chunks and source files
3
Local Ollama
Qwen generates the answer or proposed fix
4
CodePilot UI
Response, sources and safe apply controls
localhost:11434 → Ollama
Installation
Get CodePilot running in a few steps.
CodePilot currently runs with Ollama locally and installs into VS Code through the packaged VSIX release.
1
Install Ollama
Install Ollama on your system and make sure the local service is running.
2
Pull the local models
ollama pull qwen3:4b
ollama pull nomic-embed-text
3
Install the VSIX
Download the latest release, then in VS Code choose Extensions → More Actions → Install from VSIX.
4
Open CodePilot AI
Reload VS Code, open the Explorer sidebar and start using CodePilot with your project.
Ready to try it?
Install CodePilot AI in VS Code.
Download the latest VSIX release from GitHub, install it in VS Code, and run it with your local Ollama setup.