Overview
Sourcegraph 7.0 formalizes a fundamental shift in the platform's direction: positioning itself as the shared intelligence layer for both developers and AI coding agents navigating large, complex codebases. The release recognizes a critical gap in agentic development workflows—AI agents struggle with the same challenges developers face when working with large repositories, including understanding cross-repository dependencies, historical context, and architectural decisions.
Key Features
Deep Search via MCP Server AI agents can now ask semantic, cross-repository, historical, and architectural questions about enterprise codebases through Sourcegraph's Model Context Protocol (MCP) server. This enables agents to:
- Understand system connections across repository boundaries
- Reason about why code was written the way it is
- Navigate complex architectural patterns with confidence
- Access the same reliable context that developers rely on
Teams can also track MCP tool usage directly in Sourcegraph Analytics.
Enhanced Deep Search The platform now delivers more accurate results, includes image support, and integrates Code Navigation directly into Deep Search results. A new versioned Sourcegraph API is available for custom integrations.
Foundation Improvements
- Modernized Svelte UI for faster, more responsive performance
- Improved instance setup flow for faster team onboarding
- Weekly shipping cadence for Sourcegraph Cloud
Breaking Changes
Sourcegraph is removing and deprecating several features to focus on core capabilities:
- Cody Web for Deep Search customers
- Search Notebooks
- Additional features detailed in the changelog
What's Not Changing
The release emphasizes what 7.0 does not promise: perfect code generation from agents, replacement of human judgment, or a flashy UX overhaul. Instead, it focuses on infrastructure correctness and providing both developers and agents with reliable, deep code context.