Sourcegraph Cody.
Code search plus AI assistant, grounded in your full codebase index.
What it is.
Quinn Slack and Beyang Liu founded Sourcegraph in 2013 around code search. Cody is the AI layer added in 2023 — grounded in the same index Sourcegraph already maintains across customer monorepos. Used inside Uber, Lyft, Indeed, and Reddit. The self-hosted enterprise deployment runs end to end inside the customer network.
Where it fits.
Engineering organisations already paying for Sourcegraph code search. Cody is the cleanest upsell path — same index, same deployment, additional AI value on top. Strong in regulated industries that need self-hosted retrieval. The free tier offers enough to evaluate without a sales call.
- Retrieval grounded in production-grade code search index
- Self-hosted deployment story used by Uber and Lyft
- Multi-model routing including customer-deployed endpoints
- Agent UX less mature than Cursor or Windsurf
- Buyer typically needs existing Sourcegraph deployment first
Frequently asked.
Do I need Sourcegraph code search to use Cody?
No, but the value compounds when both run together. Cody alone works as a standalone assistant; with Sourcegraph index it grounds against the full codebase.
Can Cody run self-hosted?
Yes. Sourcegraph's self-hosted deployment is the reference among large engineering organisations. Customer-deployed models route inside the network without external calls.
Which model does Cody use by default?
Claude Sonnet for chat and edits; GPT-5 and Gemini available on Enterprise. Customer-deployed Llama and Mistral endpoints supported on self-hosted.
