Sentry Moves 2,500 Pages Out of CMS Using Claude Code Agents
Key Takeaways
- ▸Claude Code automated the migration of 2,500 marketing pages from CMS to Git-based Markdown, with non-technical team members successfully participating in the process
- ▸AI agent velocity was the primary driver for the migration—teams want their AI assistants to work across their entire infrastructure, making CMS silos a productivity bottleneck
- ▸Sentry achieved significant performance gains (build times from 14 minutes to under 4) and reduced code complexity by consolidating 200 templates into 3 reusable templates
Summary
Sentry successfully migrated approximately 2,500 marketing pages from their Contentful CMS to a Git-based Markdown repository using Claude Code over a four-month period. The migration was driven by the need to enable AI agents to work across their entire website infrastructure—Claude Code could rapidly update code-based pages but couldn't access CMS-locked content, creating a productivity bottleneck. The project was executed by approximately 2.5 developers, with Claude Code handling most of the actual implementation work while engineers focused on planning, directing, and reviewing.
Beyond the CMS migration, the team simultaneously migrated from Gatsby to Astro, reducing build times from 14 minutes to under 4 minutes, and consolidated 200 bespoke page templates into just 3 unified templates. The work demonstrates a broader industry trend of teams moving away from traditional CMS solutions toward more code-native approaches optimized for AI agent collaboration. Similar migrations have been documented at other companies like Cursor and Prefect.
- This represents a growing industry trend of replacing traditional CMS solutions with code-based approaches designed for AI agent collaboration
Editorial Opinion
Sentry's migration reveals an important shift in how organizations think about content infrastructure in the age of AI agents. The real insight isn't just that agents can move pages—it's that CMS silos became a productivity bottleneck once agents could automate half the work. This suggests a future where content systems are designed from the ground up for agent collaboration. Early wins like Sentry's may accelerate an industry-wide transition toward more code-native, agent-friendly architectures.
