Sourcegraph Cody
Sourcegraph Cody is an AI coding assistant for VS Code, JetBrains, Visual Studio, the web and the command line that draws context from Sourcegraph code search; it is supported on Sourcegraph Enterprise.
About this tool
Cody is Sourcegraph’s AI coding assistant. What sets it apart is the way it gathers context: it uses Sourcegraph’s search and code intelligence to find relevant snippets in the open file, the current repository and remote repositories, then sends them with your question to a large language model. Answers can therefore reflect how an API or symbol is used across a large codebase, not only the code on screen. Permissions are enforced during retrieval, so it only reads code the user is allowed to see.
Cody is available as an extension for VS Code and JetBrains IDEs, as an experimental extension for Visual Studio, inside the Sourcegraph web app and through a command-line client. It offers chat about code, auto-edit suggestions that follow your cursor and recent changes, prompts that can be saved and shared with a team, and context filters that keep chosen repositories out of chat and completions.
Sourcegraph’s documentation now describes Cody as supported on Sourcegraph Enterprise and asks interested teams to contact its sales team. Cody depends on a Sourcegraph instance both for context and as a proxy to the model provider, so self-hosted installations need internet access to reach that provider. For enterprise customers, Sourcegraph says it does not train on company data.
What you can do
Codebase context
Retrieves relevant snippets from local and remote repositories through Sourcegraph search.
Chat
Ask about, generate or edit code, adding files, symbols or repositories as context with @ mentions.
Auto-edit
Proposes changes based on cursor position and your recent edits.
Shared prompts
Save common tasks as prompts and share them across a team.
Context filters
Exclude selected repositories from chat and autocomplete.
Who is it for?
- A developer asks how an internal API is used across dozens of repositories before changing it.
- A platform team shares standard prompts for writing tests and reviewing changes.
- An organisation already running Sourcegraph adds an assistant that respects existing permissions.
Before you choose
Cody requires a Sourcegraph instance, and code snippets are sent to a third-party model provider; review this with your security team.
Availability now centres on Sourcegraph Enterprise, so check what is offered to individual developers before planning a rollout.
Frequently asked questions
Does Cody train on my code?
Sourcegraph says it does not train on enterprise customers’ data and that its model providers do not train on a customer’s codebase.
Which editors does Cody support?
VS Code, JetBrains IDEs, Visual Studio as an experimental extension, the Sourcegraph web app and a command-line client.
Can Cody run against a self-hosted Sourcegraph instance?
Yes, but the instance needs internet access because snippets are sent to a cloud model provider such as Anthropic or OpenAI.
Check the official website for current features, pricing, and terms.
Similar tools
Continue
CodingContinue was an open-source AI coding assistant for VS Code, JetBrains IDEs and the terminal. Cursor acquired the company, and the code remains available in a read-only repository.
Tabby
CodingTabby is a coding assistant that offers AI‑driven code completion, inline chat answers, and repository‑wide context.
Claude Code
CodingClaude Code is a coding assistant that lets developers edit, review, and test code using plain‑language prompts.
Codeium
CodingCodeium provides AI-assisted code completion, conversational development help and repository search through connections with supported editing environments.
Pieces
CodingPieces is a code‑assistant platform that records developer activity, organizes code snippets into a searchable archive, and provides AI‑driven chat responses that draw on that history.
Bito
CodingBito offers Governor, a layer that routes coding agents’ model requests and maps your codebase to cut AI spend, plus AI code review for pull requests.