I gave Claude Code our entire codebase. Our customers noticed. | Al Chen (Galileo)

What It Covers

Al Chen, field engineer at Galileo, demonstrates how non-engineers can clone all 15 of a company's repositories into VS Code, then use Claude Code to answer nuanced enterprise customer deployment questions that public documentation cannot address, while building customer-specific context libraries in Confluence to personalize technical responses.

Key Questions Answered

• •Multi-repo querying setup: Pull all repositories into a single parent directory in VS Code, then run Claude Code at that root level so it can traverse across all repos simultaneously. Use a Claude Code-generated 16-line shell script called "pull all" to sync the latest main branch across all 15 repos daily, replacing manual git pull commands one repo at a time.
• •Customer quirks documentation: Maintain a dedicated Confluence page per enterprise customer listing their specific infrastructure constraints — secret management tools, namespace conventions, sidecar configurations, service-to-service encryption requirements. Feed this page into a Claude Code custom command so every deployment answer is automatically tailored to that customer's environment rather than returning generic Kubernetes documentation.
• •Custom Claude Code commands: Build slash-command shortcuts (e.g., "/dpl") that chain multiple context sources — Confluence via MCP, Slack threads via MCP, and the local codebase — into a single query. This lets one command pull deployment documentation, customer quirks, and live code simultaneously, producing answers that reflect current architecture rather than potentially outdated public docs.
• •Chaos tolerance in documentation: Stop enforcing a single source-of-truth system across teams. Distribute knowledge across Confluence, Notion, and Slack freely, then use Claude Code with MCP connectors to retrieve and synthesize it on demand. The AI navigates fragmented information more effectively than humans curate it, eliminating the overhead of maintaining rigid documentation hierarchies.
• •Virtuous customer support loop: When a Slack customer thread produces a useful answer, use tools like Pylon to auto-generate a knowledge base article from that conversation. This converts individual support interactions into searchable, publishable documentation without requiring a formal PR process, creating compounding value from each resolved customer question across the entire customer base.