Sourcegraph Cody

Cody is Sourcegraph's AI coding assistant, built on the company's decade of infrastructure for universal code search and intelligence. What distinguishes Cody from many AI coding tools is its context retrieval architecture: rather than relying solely on the code visible in the current editor window, Cody can search across an entire codebase — potentially spanning millions of lines across hundreds of repositories — to find the relevant context that makes a generated suggestion accurate and consistent with the existing code style. That code intelligence foundation comes from Sourcegraph's core search product, and Cody's quality on large enterprise codebases reflects the combination of frontier LLM capabilities with genuinely deep code context.

The platform architecture separates the context retrieval layer from the LLM inference layer. When a developer asks Cody a question about their codebase, the system first searches the indexed codebase to identify relevant code snippets, then packages those snippets as context for the LLM call, then returns the model's response. Both of these steps can fail independently: the code search backend might be slow or unavailable even when the LLM provider is healthy, or the LLM API might be degraded while code search continues to return results correctly. Enterprise deployments that run Sourcegraph on-premises have an additional deployment tier that can fail independently of Sourcegraph's cloud infrastructure.

Developers experience Cody failures primarily through the IDE extensions, where chat responses begin streaming but cut off mid-sentence, or where the extension displays a loading indicator that never resolves. Autocomplete suggestions — which Cody generates continuously as a developer types — may stop appearing or become noticeably slower before stopping entirely. The context that Cody cites in its responses may be absent or less relevant than usual when the code search layer is degraded, producing responses that are technically correct language-model output but miss the repository-specific details the developer expected. The Sourcegraph web interface, where Cody chat is also available, may reflect different availability than the IDE extension if the extension uses a different API path.

Outage.gg tracks Sourcegraph Cody status through community reports from developers and engineering teams. If chat is failing, autocomplete has stopped, or responses are missing repository context, the live status page shows current incident data.