Core capabilities
Chat
Ask questions about your code, generate new code, and apply AI-suggested edits directly to your files. Cody uses your repository as context by default.
Auto-edit
Suggests code changes by analyzing cursor movements and recent edits. Can insert and delete text before and after the cursor position.
Autocomplete
Provides real-time single-line and multi-line code completions based on the code context around your cursor.
Prompts
Automate key tasks with premade and customizable prompts. Build any common query into a reusable prompt and share it with your team.
Debug code
Identify and fix errors faster. Cody can suggest fixes inline and explain bugs with full codebase context.
Context filters
Control which repositories Cody can access. Define include and exclude rules to manage context scope.
How context works
Context is what sets Cody apart from a generic LLM. When you ask Cody a question, it searches your codebase to retrieve the most relevant information before generating a response. Cody uses three methods to retrieve context:- Keyword search: Finds text matching your query, with automatic query rewriting to surface more relevant terms.
- Sourcegraph Search: Uses the native Sourcegraph Search API to query your Sourcegraph instance (managed or self-hosted). Returns relevant documents from across your entire codebase.
- Code graph: Analyzes code structure and how components are interconnected to find contextually related code.
@-mention context chips for the currently open file and repository. You can edit, remove, or add context chips at any time.
Adding context with @-mentions
You can expand or narrow Cody’s context by adding@-mentions in chat:
| Mention type | Syntax | Description |
|---|---|---|
| File | @filename | Include the full contents of a file |
| File range | @filepath:1-50 | Include a specific line range from a file |
| Symbol | @#symbolName | Include the definition of a function, class, or type |
| Repository | @repo | Search a remote repository for context |
| Web URL | @https://... | Include content from a web page |
Context scope by client
| Client | Single-repo | Multi-repo | Local context |
|---|---|---|---|
| VS Code | ✅ | ✅ | ✅ |
| JetBrains | ✅ | ✅ | ✅ |
| Visual Studio | ✅ | ✅ | ✅ |
| Cody Web | ✅ | ✅ | — |
| CLI | — | — | ✅ |
Supported clients
VS Code
Full-featured extension for Visual Studio Code. Recommended for most users.
JetBrains
Plugin for IntelliJ IDEA, PyCharm, GoLand, WebStorm, and other JetBrains IDEs.
Visual Studio
Extension for Visual Studio (experimental support).
Cody Web
Use Cody directly in the Sourcegraph web application without installing anything.
Cody CLI
Run Cody from the command line. Useful for ad-hoc exploration and scripting workflows.
Enterprise
Self-hosted or managed Sourcegraph Enterprise with Cody enabled for your organization.
Compatible with Sourcegraph products
Cody is integrated with the broader Sourcegraph platform:- Code Search: When you run a search query and open a file or repository, a Cody button takes you directly to a chat session with that code as context.
- Batch Changes: Use Cody to help write and review batch change specs.
- Code Navigation: Cody uses code graph data to surface more relevant context in its responses.
Data privacy and usage
Cody collects and uses data in the following ways:- Prompts and responses: Sourcegraph collects your prompts and responses to provide the service. For users on Sourcegraph.com, Sourcegraph may use your prompts and responses to improve the user experience but does not use any of your data to train models.
- Usage data and feedback: Sourcegraph collects usage data and feedback to improve the user experience.
Read the Cody Usage and Privacy policy for full details on how your data is handled.
Get started
Quickstart
Install Cody for VS Code and run your first chat in minutes.
Capabilities
Explore the full set of Cody features in depth.
Clients
Installation guides for all supported editors and platforms.
Code Search
Learn how Cody and Code Search work together.