Today, we’re announcing that we will end of life Sourcegraph Cody on January 31, 2027.
Cody has been an important part of Sourcegraph’s journey. We launched it at a time when AI coding assistants were just starting to change how developers work, and it helped bring AI-powered code understanding to developers around the world.
But when we decided not to evolve Cody into a coding agent two years ago, we knew that the day would come when it no longer had a place in the market.
Our mission at Sourcegraph is to bring clarity and control to the world's most complex codebases. That mission hasn't changed. If anything, I think it's become more important as AI makes it possible to create and change code faster than ever.
We can have a bigger impact by focusing on what Sourcegraph does best: helping engineering teams understand complex codebases, and making developers and the coding agents they choose more effective.
How we got here
When we introduced Cody in early 2023, AI-assisted coding looked very different. Developers were starting to use large language models for autocompletion and basic questions, but the models knew very little about the codebases they were working in.
We had spent years at Sourcegraph building the infrastructure to search and understand large codebases. Bringing that knowledge into an AI assistant was a natural next step.
And it worked. Cody was a pioneer in the RAG assistant era, bringing together AI chat and completions with Sourcegraph's code intelligence. Developers could ask questions about unfamiliar code, navigate complex systems, and get work done faster.
I'm proud of what our team built. Cody helped us understand how developers use AI. That work led us to building Amp, which eventually became an independent company. Building both Cody and Amp showed us firsthand just how much more effective AI coding tools become when they understand the codebase they’re working in.
Writing code got easier. Owning a codebase hasn’t.
Coding agents like Amp, Claude Code, Cursor, Codex, and others can now take on complex tasks and write substantial amounts of code autonomously. It’s remarkable progress, and I think we’re still early.
But as I wrote in On owning a codebase, the world's most important software sits in large, complicated codebases that have been around for years, sometimes decades. These codebases have millions of lines of code and dependencies that span teams and thousands of repositories. Just like human developers, coding agents need help understanding how it all fits together and making changes.
As I discussed in The autonomous codebase, no new model release or coding assistant alone will solve this. Even the best agent is limited by what it can find in the code. Sourcegraph is retrieval infrastructure for the agentic coding era.
Where we’re placing our bets
The future of software development won’t belong to any single coding agent. Developers will have their preferences, companies will adopt different tools (or build their own), and the frontier will keep changing.
We want Sourcegraph to work with all of them.
We're focused on building the leading platform that makes developers and their agents more effective, regardless of which tools they choose. Sourcegraph builds the deterministic code graph for your entire codebase. Why does that matter? Today's agents are incredibly capable at reasoning about code, but they often work with incomplete information.
Today, we're building around three things: helping engineering teams make faster decisions, improving the quality and accuracy of AI-generated code, and executing codebase-wide changes and security fixes with confidence.
Here's what that looks like in our products today:
- Code Search and our deterministic code graph help developers and agents find code, trace references, and understand relationships across repositories, even in the largest codebases.
- Deep Search and Code Finder are agentic search tools that help developers and agents find answers in their code, from locating a specific implementation to investigating complex issues across multiple systems.
- Our MCP, CLI, and other APIs bring Sourcegraph's code intelligence into the tools developers already use, including Claude Code, Cursor, and other coding agents.
- Agentic Batch Changes helps engineering teams plan, execute, validate, and track changes across hundreds or thousands of repositories, including dependency upgrades, migrations, and security fixes.
All of this comes back to the same thing. Give developers and agents the understanding they need to make better decisions, produce more accurate code, and make changes confidently across an entire codebase.
What this means for Cody customers
Cody will reach end of life on January 31, 2027.
I know many developers and enterprise teams still use Cody, including its IDE integrations and APIs. Some customers have built important workflows around it, and moving away from those workflows takes planning.
We're committed to working directly with impacted customers to understand how they're using Cody and help them plan their transition.
For teams using Cody for code understanding, Deep Search and our MCP integrations can support many of those workflows through the coding agents and development tools they already use. They aren't one-for-one replacements for every Cody capability, and we'll work with customers to understand their specific requirements and options.
If you're a Cody customer, please reach out to your Sourcegraph account team. We'll help you work through the transition.
Looking ahead
To everyone who used Cody, shared feedback, advocated for it inside your company, or helped build it, thank you. I'm proud of what we accomplished together. Cody reinforced something we believe deeply: AI is most effective when it understands the code it’s working with. That belief continues to guide us as we pursue our mission to bring clarity and control to the world's most complex codebases.