← All episodes

Claude Code + 15 repos: how a non-engineer answers every customer question | Al Chen

| 10 products mentioned
How I AI How I AI host
Al Chen guest
Watch on YouTube customer support automation enterprise sales enablement codebase documentation ai-powered customer success multi-repository architecture field engineering knowledge base management

Al Chen, a non-engineer on Galileo's field engineering team, demonstrates how to use Claude Code paired with a multi-repository codebase to answer complex customer questions directly from source code rather than relying on outdated documentation or engineering handoffs. By pulling 15 repositories into a single IDE workspace and querying them with AI, Chen reduces engineering team interruptions while delivering precise, current answers to enterprise customers—and shows how this workflow scales into customer success loops through documentation generation and knowledge bases.

Key takeaways
  • Load multiple repositories into a single IDE workspace rather than querying them individually; this allows AI code assistants to traverse across services and deliver contextualized answers about how interconnected systems work together.
  • Create a simple shell script (Claude can write it in one shot) to automatically pull the latest main branch across all repos daily, ensuring answers are always based on current code rather than stale documentation.
  • Build custom Claude Code commands that combine codebase queries with documentation sources (like Confluence pages on deployment or a "customer quirks" page with specific security requirements), producing highly tailored answers that beat generic docs.
  • Stop obsessing over perfect information architecture; AI can navigate messy, distributed information across Confluence, Slack, Notion, and code simultaneously, so focus on capturing knowledge rather than organizing it perfectly.
  • Turn reactive customer support into a virtuous loop by auto-generating help articles from Slack threads, publishing them to a living knowledge base, and training the entire team on answers discovered in customer conversations.
  • Hire for technical baseline in customer-facing roles or invest in Git/GitHub enablement; once the local environment is set up, querying code with AI becomes as simple as using any other chatbot.
  • Use extended reasoning prompts ("think hard, think harder") when Claude's answer feels uncertain; forcing the model to explain its reasoning and cite specific code lines catches hallucinations and often surfaces new product insights.

Recommendations (8)

Claude Code

"I can now use Claude Code to ask our entire codebase questions that are not answerable by our public documentation."

Al Chen · ▶ 0:22

VS Code
VS Code uses

"I can actually pull all of these repos into my VS Code and I can now use Claude Code to ask our entire codebase questions."

Al Chen · ▶ 5:01

Confluence
Confluence uses

"It looks at our confluence because we have a whole bunch of confluence pages about how to deploy into Kubernetes using our different images."

Al Chen · ▶ 12:14

Slack
Slack uses

"We do a lot of our customer support through Slack. We have external channels with our customers."

Al Chen · ▶ 26:37

Pylon uses

"We use a tool internally called Pylon for monitoring all our different external Slack channels."

Al Chen · ▶ 26:55

ChatGPT
ChatGPT uses

"Even using Claude Code or ChatGPT or whatever and trying to take all these different help docs."

Al Chen · ▶ 4:01

Kubernetes
Kubernetes uses

"This is all like backend images that customers have to deploy onto their Kubernetes cluster."

Al Chen · ▶ 4:41

GPT-4
GPT-4 uses

"I find myself doing this with GPT-4 which is like a powerhouse model."

How I AI · ▶ 41:21

Mentioned (2)

Notion
Notion "Throw it into confluence, throw it into notion, throw it into Slack, wherever." ▶ 0:55
Cursor
Cursor "In VS Code in cursor and whatever your IDE is, loading a project at the multi-repo level is reall..." ▶ 10:00