We spent five months running Sourcegraph's Cody across a 2,000-file monorepo and a 15-repository microservices architecture. Re-verifying pricing for this piece, we found something we should have caught a lot sooner: Cody hasn't had an individual Free or Pro tier since July 2025. Sourcegraph quietly stopped selling it to individual developers, and our own review kept describing "Free" and "Pro ($9/user/month)" as live options for over a year after that stopped being true. It's enterprise-only now, priced as "custom, contact Sourcegraph" — there's no honest single number to quote.
That correction matters more than it sounds, because Cody's actual technical differentiator is still real and still worth understanding if you're evaluating it (or budgeting for it) at the enterprise tier.
What makes Cody different
Cody sits on top of Sourcegraph's code search and intelligence platform, which means its context retrieval spans your entire organization's code — not just the current file or a single indexed project the way Cursor or Copilot work.
We tested this directly in a microservices project: the user service needed to emit an event the billing service would consume. We asked Cody how to do it. It searched across both repositories, found the existing event schema, the message queue configuration, the consumer pattern in the billing service, and the serialization helpers — then returned code that matched our existing patterns exactly, without us curating any of that context by hand. We haven't seen another AI coding tool do cross-repository reasoning at that depth.
The tradeoff is latency. Because Cody searches before it answers, responses are noticeably slower than Cursor or Copilot — fine for a complex, codebase-spanning question, unnecessary overhead for a simple one.
Where it's worth it, and where it isn't
Autocomplete is competent (roughly Copilot-tier, behind Cursor's Tab completions) but gets meaningfully better when your code needs to match existing conventions — in a Go codebase with 12 existing repository implementations, Cody's completions matched our interface pattern, error handling, and logging format precisely, where Cursor's were syntactically correct but generic.
Multi-LLM support (Claude, GPT-4o, Gemini, and others, with enterprise model policies and private deployment options like Bedrock or Azure OpenAI) is genuinely useful for regulated environments that need to control where code gets processed.
None of this changes the bottom line on price: if you work in a single small-to-medium project, the cross-repo intelligence doesn't earn its keep, and there's no cheap way in anymore to find out — the on-ramp for individuals is gone.
The takeaway
Cody is a specialized tool for a specific problem: understanding code relationships across many repositories at organizational scale. If that's your problem, it does it better than anything else we've tested. If it isn't, the free/Pro path that used to let you find that out for yourself no longer exists — verify current pricing directly with Sourcegraph before you plan around a number, including any number in an older review (ours included, before this correction).
Full review, cross-repo examples, and the autocomplete comparison: https://devtoolsreview.com/reviews/cody-review/
Top comments (0)