What it does
Cody is Sourcegraph's AI coding assistant. The docs say it pairs current language models with context from your own code, pulled through Sourcegraph's Search API from local and remote codebases, so answers can draw on APIs, symbols and usage patterns from across a whole codebase. It is offered as extensions for VS Code and JetBrains, an experimental extension for Visual Studio, a web app inside Sourcegraph, and a command-line client. The docs state that Cody is supported on Sourcegraph Enterprise, and they point readers to the Sourcegraph team to try Cody Enterprise. It connects to code hosts such as GitHub and GitLab. The main features listed are chat with @-mentions for adding files, symbols or remote repositories as context, auto-edit suggestions based on cursor movement and recent changes, premade and customizable prompts, debugging help, and context filters that keep chosen repositories out of chat and autocomplete. On data, the docs say Sourcegraph collects prompts and responses to provide the service, and that it does not use your data to train models.
Capabilities
- Chat that starts with your open file and repository, with @-mentions for more context
- Auto-edit suggestions based on cursor movement and recent changes
- Premade and customizable prompts that can be saved and shared with a team
- Code completions and code edits inside the editor
- Debugging help for finding and fixing errors
- Context filters to exclude selected repositories
When to use it
- Asking questions about a large codebase from inside your IDE
- Writing or fixing code with context from several repositories
- Sharing standard prompts across a development team
- Keeping selected repositories out of what an AI assistant can see
