Most AI coding assistants work from whatever happens to be open in your editor. Sourcegraph Cody starts from a different premise: the code you need to reason about usually lives somewhere else in the repository, or in another repository entirely. Cody is built on Sourcegraph's code intelligence platform, and it uses that platform's search to pull relevant context before answering. That design choice shapes where it helps and where a lighter-weight tool would do just as well.
Three situations show the fit clearly:
- Onboarding onto an unfamiliar codebase. A new engineer can ask why a service behaves a certain way, and Cody's chat can reference the actual files and functions involved rather than guessing from a generic pattern. Because context is retrieved from the indexed code, answers point back to real definitions instead of plausible-sounding invention.
- Working across many repositories at once. In organizations where a feature touches several services, Cody's use of Sourcegraph's Search API to pull from local and remote codebases means a question can span more than the file you happen to have open. This is the scenario the tool is genuinely differentiated for.
- Large-scale migrations and refactors. Sourcegraph frames its enterprise tooling around completing large refactors and migrations in hours rather than days, pairing AI edits with Batch Changes for repetitive, repo-spanning modifications. If you routinely make the same change across dozens of files or repos, this is the core use case.
How the context model actually works
The feature that everything else depends on is context retrieval. Cody uses Sourcegraph's Search API to find relevant code in local and remote repositories and feed it to the model. In practice this matters because the quality of an AI answer is capped by the quality of the context it was given. A tool that only sees the current file will confidently write code that duplicates an existing helper or ignores a project convention two directories over. By searching the wider codebase first, Cody has a better chance of grounding its response in how your project already does things.
That retrieval feeds the visible features. Chat handles the open-ended work: explaining a function, sketching an implementation, or reasoning about a bug, with codebase context attached. Code completions and edits suggest changes based on your cursor position and recent edits, which is the day-to-day autocomplete loop. Prompts let you save and share reusable instructions for recurring tasks, so a team can standardize things like "write a test in our style" instead of everyone re-typing the same request. There is also dedicated support for debugging workflows around identifying and fixing errors.
One underrated control is context filters, which let you exclude specific repositories from results. In a monorepo or a large org, not all code should be pulled into a prompt, and being able to scope what the assistant can see is both a relevance and a governance feature.
Where it runs and what it connects to
Cody is available in VS Code and JetBrains IDEs, with Visual Studio support marked experimental, plus a web app and a CLI. The CLI matters more than it sounds: it lets the same context-aware assistance run in scripts and terminals, not just inside an editor. On the code host side, Sourcegraph connects to GitHub and GitLab, which is how it indexes the remote repositories that feed context. If your source lives outside those hosts, confirm coverage before committing.
Who should look at this seriously
The honest answer is that Cody earns its keep in proportion to codebase size and complexity. A solo developer on a small project will get most of the value from any competent completion tool, and the codebase-search machinery is overhead they won't use. The teams that benefit are those working in large, multi-repository systems where the hard part of a change is understanding what already exists. Engineering organizations that also want search, refactoring, and analytics over their code, rather than an editor plugin in isolation, are the intended audience, since Cody sits inside the broader Sourcegraph platform. You can compare it against other tools in the AI code generator category to see how that platform-first approach stacks up.
Pricing, and a caveat about tiers
Sourcegraph's pricing has consolidated around an Enterprise plan, publicly listed with a starting price of $16K and a credit-based model: credits are included per user, support org-wide pooling, do not expire monthly, and roll over on renewal. The enterprise package bundles code search and navigation, AI-powered Deep Search, Batch Changes, Insights and Monitoring, and full MCP server, API, and CLI access, along with a single-tenant cloud deployment option and enterprise-grade security and admin controls. Support is described as 24x5 with upgrade options, and volume pricing is available.
A note on the free tier: Cody has historically offered a free plan for individual developers, and the tool is still categorized as freemium. However, the current public pages emphasize the Enterprise offering and do not lay out individual free or Pro tier details, so treat any specific free or Pro limits as something to confirm directly with Sourcegraph rather than assume. The $16K figure is a starting point for enterprise, not a per-seat rate; actual cost scales with team size.
| Aspect | What is publicly documented |
|---|---|
| Enterprise starting price | $16K, scales with team size, volume pricing available |
| AI usage model | Credits per user, org-wide pooling, roll over on renewal |
| Deployment | Single-tenant cloud option |
| Free / Pro individual tiers | Not detailed on current pricing page; confirm with vendor |
Data handling and the enterprise controls
For teams whose main hesitation about AI coding tools is data exposure, Sourcegraph documents a fairly specific position. Its enterprise materials cite SOC 2 Type 2 compliance, GDPR and CCPA compliance, audit logs, and a stance that its AI models do not retain data from requests and that customer data is not used for model training. It also lists uncapped IP indemnity for generated code, public code guardrails intended to prevent open-source license violations, and full code ownership retained by the customer. Combined with the context filters mentioned earlier, this gives a security team concrete things to evaluate rather than vague reassurances.
The tradeoffs worth weighing
The biggest limitation is that Cody's strengths and its cost are entangled with the Sourcegraph platform. The context advantage comes from indexing your repositories, which means real value depends on getting that indexing set up and on your code living in supported hosts like GitHub or GitLab. For a small team, that is more setup than a drop-in editor plugin, and the enterprise-oriented pricing puts the fullest version out of reach of hobby projects.
Two more honest caveats: the current public pages are light on individual free and Pro tier specifics, so budget-conscious solo users can't easily confirm limits without contacting the vendor, and Visual Studio support is still experimental, so JetBrains and VS Code are the mature paths. None of these are dealbreakers for the target audience, but they do reinforce that this is a tool aimed at engineering organizations rather than casual use.
Verdict
Sourcegraph Cody is a strong choice when the difficulty of your work is understanding a large, interconnected codebase, not just typing code faster. Its retrieval-based context, cross-repository reach, prompts, and Batch Changes integration are aimed squarely at teams doing serious work across many services, and its documented security posture gives enterprises a reason to trust it with proprietary source. If you are an individual on a small project, the platform overhead and enterprise pricing likely outweigh the benefit, and a lighter completion tool will serve you better. For everyone else, the deciding question is simple: does your hardest daily problem involve navigating code you did not write? If so, Cody is built for exactly that. Browse more options on the tools directory or read comparisons on the blog before deciding.
Common questions about Sourcegraph Cody
How is Cody different from a standard AI autocomplete tool?
Cody retrieves context from your local and remote repositories using Sourcegraph's Search API before responding, so its chat and edits can reference code beyond the file you have open. Standard autocomplete tools typically work only from nearby code, which is why Cody's advantage grows with codebase size.
Which editors and platforms does Cody support?
It runs in VS Code and JetBrains IDEs, with Visual Studio support marked experimental, plus a web app and a CLI. On the code host side, it connects to GitHub and GitLab.
What does Sourcegraph cost?
The publicly listed Enterprise plan starts at $16K and uses a per-user credit model with org-wide pooling and credits that roll over on renewal; pricing scales with team size and volume pricing is available. Details for individual free or Pro tiers are not laid out on the current pricing page, so confirm those directly with Sourcegraph.
How does Cody handle my source code and data?
Sourcegraph's enterprise documentation states that its AI models do not retain data from requests and that customer data is not used for model training. It also cites SOC 2 Type 2, GDPR and CCPA compliance, audit logs, uncapped IP indemnity for generated code, public code guardrails, and context filters to exclude selected repositories.
Is Cody a good fit for a solo developer?
Its differentiation depends on indexing a large, multi-repository codebase, so a solo developer on a small project gets less from the context features and faces platform setup and enterprise-oriented pricing. It is aimed at engineering teams working in complex code.
