Sourcegraph Cody

What it actually does

The thing to know first: Sourcegraph discontinued Cody Free, Cody Pro and Cody in the Enterprise Starter plan on 23 July 2025. The company's own announcement, published 25 June 2025, said "as of July 23, 2025, the following plans will no longer be available" and that "existing users on these plans will retain access to Cody until that date." Paying individual users were cut off and steered toward a different product, Amp. Anything you read about Cody's $9 Pro plan, including the vendor copy still circulating, describes something you can no longer buy.

What survives is Cody Enterprise. The technology itself is real and the idea was good: it indexes your whole codebase using Sourcegraph's code search infrastructure, so chat and autocomplete answer with context from across repositories instead of just the open file. It works in VS Code and JetBrains, supports MCP, and lets you bring your own model rather than locking you to one vendor's, which is genuinely useful for teams with privacy constraints or on-premise requirements.

What it costs

There is no published price. The individual tiers are gone, so the only route is through enterprise sales, with a reported minimum annual contract around $16,000.

That is the entire pricing story, and it is the reason for the 1.0. A solo developer cannot buy this. A five-person startup cannot realistically buy this. The old free plan (9 chats and 500 autocompletes a month) no longer exists as an evaluation path, so there is no way to try before you enter a sales conversation about a five-figure commitment.

Where it falls down

The discontinuation is the failure. Removing a paid tier from under existing customers, then pointing them at a replacement product rather than an equivalent plan, is the kind of vendor behavior that should make anyone think twice before depending on the rest of the lineup.

On the product itself, the complaints that exist are pointed. Developers who tried it report output that ignored the existing patterns in their codebase, one describing the results as random crap. Users also report no memory across sessions, so closing the tab discards everything you fed it and you re-prime the context every time you sit down. For a tool whose selling point is context, that is a strange gap.

The evidence problem is as bad as it gets on this site. The only rating we could find is 2.8 out of 5 from two G2 reviews, reported by The AI Agent Index. Two reviews. Cody is effectively invisible in the developer community next to Copilot and Cursor, so a buyer considering a five-figure contract has almost no independent experience to check against the sales pitch.

The positives are real but narrow: multi-repository context through Sourcegraph's search is strong, and for a large security-conscious engineering organization already running Sourcegraph, the enterprise deployment options make sense.

Who should buy it

Large engineering organizations that already pay Sourcegraph, have the code search index in place, and need an assistant that can run against on-premise or self-chosen models. For them this is a line item on an existing relationship rather than a new bet, and the codebase context is worth something.

Individual developers should buy nothing here, because there is nothing to buy. If you were a Cody Pro subscriber who got cut off, your realistic options are Copilot or Cursor, both of which have self-serve pricing and enormous user bases you can actually research. Small teams are in the same position.

We rate this 1.0 not because the technology was bad but because the thing a normal reader of a review site could purchase was taken away, the replacement is a different product, and what remains is priced and evidenced in a way that makes independent evaluation impossible.