Chat
Discover how Cody's Chat feature provides contextually-aware answers and code suggestions, enhancing your coding workflow.
Cody's Chat feature allows you to interact with AI to ask questions about your code, generate new code snippets, and refactor existing code. It leverages context from your open files and repository to provide accurate and relevant responses.
Versatile Assistance
Get help with code understanding, generation, and editing
Context-Aware Responses
Answers are informed by your codebase for better accuracy
How Chat Works
Cody's Chat uses a sophisticated approach to provide you with the best possible assistance:
-
Codebase Search
Cody searches your codebase using Sourcegraph's native search and keyword search to find relevant context.
-
Contextual Understanding
By understanding the context, Cody provides informed responses specific to your project.
-
Informative Responses
Cody indicates which code files it has analyzed to generate its responses, ensuring transparency.
Key Features
Default Context
Automatic context chips for opened file and repository in new chats.
Custom Context
Add context using @-mentions: repositories, files, symbols, and line ranges.
OpenCtx Support
Experimental support for fetching context from web URLs and non-code artifacts.
Offline Mode
Experimental offline chat via Ollama, no Sourcegraph account needed.
LLM Selection
Choose between different LLMs for chat, optimized for speed or accuracy.
Smart Apply
Dynamically insert code from chat into your files with intelligent placement.
Chat History
View, export, and clear your chat history for session management.
Quick Prompts
Ready-to-use prompts for common coding actions.
Code Generation Example
See how Cody can write code for you. In VS Code, ask Cody to "write a function that sorts an array in ascending order."
Write a function that sorts an array in ascending order
Cody provides code suggestions directly in the chat window, offering options to:
Live Chat Demo
Interact with the embedded chat demo below to experience Cody's chat capabilities firsthand.
Try asking questions about code or requesting code generation.

Improving Cody's Answers
If Cody's initial response isn't quite right, refine your approach:
- Use Public Knowledge Only: Limit context to general model knowledge for broader queries.
- Focus on Current File: Restrict context to the open file for file-specific questions.
- Add Specific Context: Utilize @-mentions to include relevant files, symbols, or external resources for more precise answers.