Sourcegraph Cody Review: Codebase-Wide AI Help for Big Teams

Cody is the coding assistant for organisations whose code is too big for one developer's head. Built on Sourcegraph's code search, it answers questions and writes code with context from across repositories, which makes it unusually good at legacy systems, cross-service changes and onboarding. The catch is access: individuals can no longer buy it at all, and enterprises face a large annual minimum. For the companies it serves, though, that codebase-wide context is genuinely hard to get elsewhere.

What it is

Cody is Sourcegraph's AI coding assistant, sold as part of the Sourcegraph Enterprise platform rather than as a standalone product. It lives in editors such as VS Code and JetBrains, on the web and in the command line, and it draws on Sourcegraph's code intelligence: an index of your repositories, their symbols, references and history. Instead of guessing from the file you have open, Cody can ground answers in code from across services, which suits large, long-lived codebases. Individual Free and Pro plans were discontinued in July 2025, with new signups stopping that June; individual developers were pointed at Amp, Sourcegraph's separate agentic coding tool, while Cody continues for enterprise customers.

Key features

Codebase-aware chat is the core: ask how a feature works, where a value is set or why a service behaves a certain way, and Cody answers with references to real code across repositories. Completions and edits work in the editor, with multi-file changes for refactors that span services. Reusable prompts let teams standardise common tasks, such as writing tests or documenting an API, so every developer gets the same quality of help. Code search across all repositories, including remote hosts, means context is not limited to what is checked out locally. Batch Changes extend this to organisation-wide modifications, applying a change across many repositories with review, and enterprise controls cover permissions, self-hosting and audit requirements.

Pricing

Cody has no current individual pricing. Cody Free and Cody Pro were discontinued on 23 July 2025, and there is no free tier or self-serve paid tier to buy today. Cody is available only inside Sourcegraph Enterprise, which Sourcegraph lists as starting at around 16,000 dollars a year on an annual contract, with the final figure set by seat count, deployment type and bundled AI usage through sales. Some third-party trackers quote roughly 59 dollars per user per month for Cody Enterprise, but Sourcegraph does not publish that number itself, so treat it as unverified. Evaluations typically happen through a guided demo rather than a free trial.

What users like

Users commonly praise Cody for codebase-wide context, with reviewers highlighting answers that reference the right files across services instead of generic explanations. Large-team users value consistency: shared prompts and whole-repo awareness help new developers onboard and help reviewers understand unfamiliar code faster. Enterprise buyers also appreciate the governance story — permissions, self-hosting options and audit support — which matters when code cannot leave the company's environment. For legacy and monorepo work in particular, reviewers describe Cody as more grounded than assistants that only see open files.

What users complain about

The most common complaints are about access and price: individuals and small teams can no longer buy Cody at all, and the enterprise minimum puts it out of reach for most smaller organisations. Some users find the product story confusing, with Cody's repositioning and the separate Amp tool making it unclear what they are actually being sold. Others note that setup and indexing take effort on very large codebases, and that answer quality still depends on the underlying models and on how well the codebase is organised. A smaller theme is the lack of a self-serve trial for evaluation.

Limitations to know

Cody can still produce wrong, insecure or outdated code, and its answers are only as good as the code it can see and the models behind it, so review and test everything before merging. Codebase-wide context does not mean it understands business rules, undocumented behaviour or code it cannot index, and it may confidently explain code that is itself buggy. Because Cody is enterprise-only, teams should confirm entitlement, deployment options and data handling with Sourcegraph sales before planning around it, and individuals should evaluate Amp or another assistant instead.

Who it is best for

Cody suits large engineering organisations with many repositories, long-lived systems and strict governance needs, where whole-codebase context genuinely changes what an assistant can do. It fits platform teams, legacy modernisation work and organisations already using Sourcegraph code search. It is not for individual developers, freelancers or small teams, who cannot buy it and should look at Amp, open-source assistants or subscription rivals instead.

Verdict

Cody is a strong assistant trapped behind an enterprise gate. Its codebase-wide grounding is a real advantage for large organisations, and its governance features suit regulated environments. But the discontinued individual tiers and the large annual minimum mean most developers will never use it, which limits its recommendation to the enterprises it is actually sold to. Our editorial score: 7.8/10 — our own judgement, not an aggregate of user ratings.

See the Sourcegraph Cody directory entry · All AI tools · All guides