Skip to content
Agents tracked: 258 Downloads (7d): 219M up 6.1% GitHub stars: 5.5M VS Code installs: 148M Releases (7d): 293 Agent pull requests (last week): 937K Updated Oct 7, 2026
Case study LG CNS

How LG CNS migrated a 20-year-old system with a Claude Code build pipeline

LG CNS built an orchestration layer around Claude Code, which it calls a Build Factory, and used it to modernize a construction company's 20-year-old project management system. As reported by Anthropic, this took roughly half the cost of a conventional rebuild.

The problem

LG CNS's customers needed to modernize old mission-critical systems, but cost, schedule and quality risks often kept those projects from starting. One client's 20-year-old project management system would have needed about twice the available time and budget.

How they did it

  1. Described the project goals to Claude and asked it to recommend an approach.
  2. Broke the work into small, clearly scoped tasks defined in YAML.
  3. Passed outputs between stages as files on disk, which served as persistent memory.
  4. Used Claude Opus 4.6 for architecture design, legacy analysis and code generation.
  5. Added thin layers for validation, context scoping and quality checks.

Results

  • As reported by Anthropic: 2,888 of 2,913 APIs converted, 'a 99.1% completion rate'.
  • The migration was delivered 'in seven months, at roughly 50% of the cost of a conventional rebuild'.
  • 1,340 screens moved from a proprietary legacy frontend platform to React.
  • Claude Code became the standard development approach across the 200-engineer Build Center.

As reported by the source (Anthropic customer story); AgentGid did not measure these figures.

Takeaway. Break a large migration into small, explicitly defined tasks and pass results between stages through files, so the agent's context stays focused.
AgentGid's take

This suits an engineering organization with enough people to build and maintain its own orchestration layer: task definitions in YAML, file-based handoffs, and validation checks. The 99.1% completion and 50% cost savings are as reported by Anthropic, and they don't cover the review effort for the remaining APIs. To experiment cheaper, Pi is Free (OSS), though you pay your model provider.

The agent used here

Similar use cases

Guides