Quality assurance · Production

Claude Code: Anthropic's Agentic Coding Tool in Your Terminal

The problem

Developers lacked a composable, pay-as-you-go agentic coding tool with raw model access that could be embedded in larger workflows, rather than wrapped in a polished IDE with hidden scaffolding.

Workflow diagram · grounded in source
1
Terminal invocation
trigger
“Claude Code, it runs in your terminal. Because it runs in the terminal, it has access to a bunch of stuff that you just don't get if you're running on the web or on desktop or whatever. So it can run bash commands, it can see all of the …”
2
Parallel instance scaling
ai_action
“if you have like a thousand Lint violations and you want to start a thousand instances of Claude and have it fix each one and then make a PR, then ClaudeCode is a pretty good tool”
3
Context compaction
ai_action
“we actually just did the simplest thing, which is ask Claude to summarize the, you know, the previous messages and just return that”
4
GitHub Action lint trigger
trigger
“we have this GitHub action that runs. And the GitHub action invokes quad code with a local slash command. And the slash command is lint. So it just runs a linter using quad.”
5
Semantic linting
ai_action
“it'll check for spelling mistakes, but also it checks that code matches comments. It also checks that, you know, we use a particular library for, you know, a specific command. So it checks for network fetches instead of the built-in libr…”
6
GitHub MCP commit
integration
“use the GitHub MCP server in order to commit the changes back to the PR”
Reported outcome

Claude Code achieved rapid internal adoption at Anthropic with vertically rising daily active users; approximately 80–90% of Claude Code itself is now written by Claude Code, and average external user spend is $6 per day.

Reported metrics
Claude Code-written share of Claude Code codebase80, 90%
Average cost per active user per day$6 per day
Estimated engineer productivity improvement50, 70% more productive
maximum single-day spend by power users at Anthropic>$1,000 in one day
Show all 5 reported metrics
Claude Code-written share of Claude Code codebase80, 90%
average cost per active user per day$6 per day
estimated engineer productivity improvement50, 70% more productive
maximum single-day spend by power users at Anthropic>$1,000 in one day
cost comparison: Claude Code vs Cursor$6/day on Claude Code compared to $20/mo for Cursor
Reported stack
Claude CodebashMCPGitHub MCP serverGitHub actionTmuxCommanderJSPuppeteerGitHub
Source
https://www.latent.space/p/claude-code
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Claude Code achieved rapid internal adoption at Anthropic with vertically rising daily active users; approximately 80–90% of Claude Code itself is now written by Claude Code, and average external user spend is $6 per…

What tools did this team use?

Claude Code, bash, MCP, GitHub MCP server, GitHub action, Tmux, CommanderJS, Puppeteer, GitHub.

What results were reported?

Claude Code-written share of Claude Code codebase: 80, 90%; Average cost per active user per day: $6 per day; Estimated engineer productivity improvement: 50, 70% more productive; maximum single-day spend by power users at Anthropic: >$1,000 in one day (source-reported, not independently verified).

How is this quality assurance AI workflow structured?

Terminal invocation → Parallel instance scaling → Context compaction → GitHub Action lint trigger → Semantic linting → GitHub MCP commit.

Related quality assurance cases