What Happened
Al Chen, a field engineer at Galileo (an AI observability platform), built a customer support system using Claude Code that queries all 15 of Galileo's repositories simultaneously. Without a formal engineering background, he combined repo context with Confluence documentation and Slack history via MCP integrations. A 16-line script written by Claude Code pulls the latest main branch across all repos to keep context current. The result: enterprise customer questions that previously required engineering escalation are now answered directly by customer-facing staff.
Why It Matters
This case demonstrates a concrete productivity pattern for indie developers and small teams: using Claude Code as a codebase-aware knowledge layer, not just a code generator. Key mechanics include:
- Opening VS Code at the multi-repo root level so Claude Code sees all 15 projects as one context window
- A "customer quirks" file per enterprise client that encodes deployment-specific edge cases
- MCP connectors linking Confluence docs and Slack threads to the same query interface
- Treating live code as the authoritative source of truth over documentation, which drifts
For SMEs with small support teams, this pattern can reduce engineering interruptions to near-zero while improving answer accuracy, since code reflects actual behavior rather than outdated wikis.
Asia-Pacific Angle
Chinese and Southeast Asian SaaS companies expanding into enterprise markets face a specific challenge: English-language enterprise customers expect deep technical support, but engineering bandwidth is constrained and time zones create response delays. This multi-repo Claude Code pattern is directly replicable using open-weight models like Qwen2.5-Coder for on-premise deployments where IP sensitivity prevents sending proprietary code to Anthropic's API. Teams in Singapore, Vietnam, or Shenzhen serving Japanese or Australian enterprise clients can pre-load customer-specific quirk files in local language context and query in English output, bridging both technical and linguistic gaps without hiring additional engineers.
Action Item This Week
Write a shell script (or ask Claude Code to write it) that runs git pull origin main across every repository in your product, then open the parent directory in VS Code with Claude Code enabled. Ask one real customer question against the full codebase context and measure the answer quality versus your current documentation.