Sourcegraph's Cody docs still stand while its blog moves on

The setup

Cody is Sourcegraph's AI coding assistant, built around indexing and searching a company's codebase rather than answering only from the file open in an editor. Sourcegraph's own documentation, retrieved 16 September 2026, states the tool is supported on Sourcegraph Enterprise and works through extensions for VS Code, JetBrains, Visual Studio, or the Sourcegraph web app once a codehost such as GitHub or GitLab is connected. Reaching the feature therefore requires an existing Sourcegraph Enterprise deployment, not just installing an editor plugin; the connection to an indexed codebase is the setup step the extension itself depends on.

What the documents show

The documentation states plainly what makes Cody's context different from a general assistant: it uses Sourcegraph's own advanced Search API to pull context from both local and remote codebases, so a developer can draw on APIs, symbols and usage patterns from across an entire codebase rather than a single open file. On scoping, the same page names a specific control, Context Filters, which let Cody ignore selected repositories from chat and autocomplete results, Sourcegraph's documented way for a company to exclude repositories rather than something a user configures per query. On data use, the documentation states that for individuals using Cody through Sourcegraph.com, none of that data is used to train models, a vendor statement about its own handling, not an outside audit.

The friction

Checking Sourcegraph's own product blog against the Cody documentation surfaces a gap the docs page does not mention: the roughly twenty most recent posts, as retrieved 16 September 2026, describe products including Deep Search, Code Finder, an MCP server, and Agentic Batch Changes for edits across thousands of repositories, and Cody does not appear by name in any of them. That is a factual observation about where the company's current public attention sits, not a documented statement that Cody is being retired, and it should not be read as more than that; the Cody documentation page itself carries no deprecation notice.

What changed in the work

What the documentation directly supports is unchanged since Cody's introduction: context comes from an indexed, permissioned view of a codebase rather than an ad hoc paste into a chat window, and a company can narrow that scope with Context Filters. What is genuinely uncertain from these two sources alone is whether Cody remains the vendor's primary investment or has become a maintained but quieter product line next to newer agent tools; a team relying on it for Context Filters and Enterprise integration should treat that uncertainty as worth a direct question to Sourcegraph, not an assumption either way.

• Is this Sourcegraph Enterprise deployment's Context Filters list actually excluding the repositories it is meant to?
• Has the vendor said anything, outside the blog's recent silence, about Cody's ongoing support relative to its newer agent products?
• Does the no-training statement in the documentation match the data-handling terms in this organization's actual Sourcegraph contract?

A product's own documentation can be accurate and current while its publisher's attention has visibly moved elsewhere, and the two are worth checking separately rather than assuming one implies the other.