Sourcegraph Cody

Cody is Sourcegraph's AI coding assistant that combines large language models with deep, comprehensive codebase awareness. Unlike generic AI chat tools that only see the single file you currently have open in your editor, Cody connects directly to Sourcegraph's enterprise code search infrastructure to pull context from your entire repository, including local files, remote codebases, and even code spread across multiple repositories in your organization. This means it actually understands your specific APIs, internal symbols, naming conventions, and real usage patterns when answering questions or generating code, rather than just guessing based on common open source conventions or generic programming patterns.

The core capability is a developer chat interface embedded in your IDE where you can ask questions about your codebase, generate new code that follows your existing patterns, or request edits to existing files with full awareness of how those changes will interact with the rest of your system. Cody maintains awareness of your currently open files while simultaneously having access to the broader repository context through Sourcegraph's search API. When you ask how a particular function works, where a certain pattern is used throughout the codebase, or what the expected behavior of an internal API is, Cody can search across the entire codebase to give you accurate, grounded answers rather than hallucinating plausible sounding responses that don't match your actual code.

Beyond conversational chat, Cody includes an auto edit feature that watches your cursor movements and editing patterns, then suggests contextual code modifications as you work. It's like having a pair programmer sitting next to you who can see what you're doing and offer relevant, codebase aware suggestions without waiting to be asked. The system notices when you're about to implement something that exists elsewhere in the codebase and can pull in the right patterns automatically. There's also a prompts system for automating common development tasks through pre built templates that you can customize for your specific workflow. If you find yourself repeatedly performing the same kind of refactoring, code generation, test writing, or documentation task, you can build a reusable prompt that encapsulates your approach.

The context engine is fundamentally what makes Cody different from simpler AI coding tools that treat every conversation as isolated. It integrates directly with Sourcegraph's search API, which is the same technology that powers enterprise code search at major technology companies managing millions of lines of code across thousands of repositories. This lets Cody pull relevant information from both your local development environment and remote repositories you have access to, giving it a comprehensive view of your codebase that matches what a senior engineer with years of context would have. Context filters let you exclude specific repositories from results if you need to keep certain proprietary or sensitive code out of the AI's view for compliance or security reasons.

IDE integration covers the major development environments where professional developers spend their working hours. You can run Cody in VS Code with full feature support, JetBrains IDEs like IntelliJ IDEA, PyCharm, WebStorm, and the rest of the JetBrains family, Visual Studio in an experimental mode for the Windows development crowd, or through a web interface and command line for specialized workflows. This flexibility means Cody fits into most existing development setups without forcing you to change your preferred editor or adopt new tooling. The experience is kept consistent across platforms, though some features may be more polished in VS Code where the extension ecosystem is most mature and Sourcegraph has invested the most engineering effort.

Cody runs on Sourcegraph's enterprise infrastructure with clear data handling policies, which addresses the privacy and security concerns that have blocked AI coding tool adoption at many companies working on proprietary software. Sourcegraph explicitly states that for Sourcegraph.com users, they don't use any of your code to train models. Your proprietary code stays private and is only read to answer your specific questions within your authenticated session. This matters enormously for teams working on trade secret software, regulated industries, or any codebase where leaking snippets into a public training dataset would be unacceptable. The enterprise focus extends to deployment options, with self hosted configurations available for organizations that need to keep everything within their own infrastructure.

The pricing model is enterprise focused with dedicated support packages for companies running Sourcegraph Enterprise. This positions Cody for engineering teams that are already invested in code search capabilities and want AI assistance built on top of that foundation rather than bolted on as an afterthought. If your organization already uses Sourcegraph for code navigation, search, and code intelligence, adding Cody is a natural extension that leverages the infrastructure you've already deployed. For individual developers or small teams not already paying for Sourcegraph, the value proposition depends on whether the superior codebase awareness and enterprise grade privacy guarantees justify the pricing compared to lighter weight alternatives that work well enough for smaller, simpler codebases.