Cody is Sourcegraph's AI code assistant, built on the same code graph technology that powers enterprise code search — giving it the repository context that generic AI assistants lack when working on large, multi-repo codebases.
The Problem
ChatGPT and Copilot work well on isolated snippets. They cannot tell you why a function was deprecated three releases ago, trace a data flow across ten microservices, or explain how a configuration change in one repository will affect a dependent service. Their context is limited to what you paste in, not the full system that actually exists in your organization.
How Cody Solves It
Cody indexes your entire codebase with Sourcegraph's code intelligence graph. It understands relationships between files, uses precise code navigation to trace call chains, and pulls relevant context automatically rather than asking you to attach files manually. Ask it to explain a function — it reads the callers, the types, and the historical changes that shaped it.
Key Features
- Full-codebase context via Sourcegraph code graph, not just currently open files
- Chat, inline completion, and edit commands (fix, generate, explain, test)
- Multi-repo support for monorepos and distributed microservice architectures
- Works with Claude, GPT-4, and Gemini; model switchable at the organization level
- VS Code, JetBrains, and Neovim extensions
Who It's For
Cody is best for engineering teams with large, multi-repository codebases who need AI assistance that understands the full system — not just the currently open file — and for organizations already invested in Sourcegraph for code search and navigation.
Compared to GitHub Copilot
Unlike GitHub Copilot, Cody uses Sourcegraph's code graph to understand relationships across your entire repository — it traces call chains, reads callers and type definitions, and provides context-aware answers that go far beyond the currently open file.
