How Rakuten ran a seven-hour autonomous vLLM change with Claude Code
Rakuten engineers use Claude Code across the development lifecycle. In one test, it implemented a method in the open-source vLLM library in a single unattended run.
The problem
Rakuten wanted to scale AI-assisted development across thousands of developers without lowering quality. Earlier tools needed frequent human guidance and struggled with large, multi-language codebases.
How they did it
- Used Claude Code for unit tests, API mocks, components, bug fixes, documentation and code review.
- Ran several tasks in parallel by handing most of them to Claude Code while engineers worked on one.
- Had new team members use Claude to understand complex codebases and past architectural decisions.
- As a test, gave Claude Code an activation vector extraction task in vLLM, a library with 12.5 million lines of code.
Results
- As reported by Anthropic: 'Claude Code finished the entire job in seven hours of autonomous work in a single run.'
- The implementation reached '99.9% numerical accuracy compared to the reference method'.
- Time to market for new features fell 'from 24 working days to 5 days—a 79% reduction'.
As reported by the source (Anthropic customer story); AgentGid did not measure these figures.
This suits teams with a mature codebase, good tests, and engineers who can write tight specs and review large diffs, since the unattended run only helps if someone can verify the output. Note that the 7-hour, 99.9% accuracy, and 79% figures are vendor-reported by Anthropic, and Claude Code needs a paid plan (from $20/mo). Pi (Free, OSS) is a cheaper terminal alternative, though you pay model costs.
The agent used here
Similar use cases
16 parallel Claude agents wrote a C compiler that builds the Linux kernel
An Anthropic researcher ran a team of Claude agents in a loop on a shared repository to write a Rust-based C compiler from scratch. The write-up and the code are both public.
Over nearly 2,000 Claude Code sessions and $20,000 in API costs, the agents produced a 100,000-line compiler.
Turn a Figma frame into code by connecting the Figma MCP server
Figma's hosted MCP server gives a coding agent the variables, components and layout data of a frame you link to, so the code it writes matches the design.
The agent gets design context such as variables, components and layout for the selected frame.
JustHTML: a Python HTML5 parser built with coding agents that passes html5lib tests
A developer spent his off-hours using Copilot agent mode with several models to build a dependency-free HTML5 parser, and wrote up every stage, including a full restart.
JustHTML is about 3,000 lines of Python with 8,500+ tests passing.
Guides
Best AI Agents in 2026: Live Rankings by Category
The leading AI agents for coding, design, research, data, marketing, sales, support, HR, finance, legal, DevOps and security, ranked by public usage data that refreshes daily, with prices and free options.
What Is an AI Agent? A Practical Explanation With Real Examples
What makes software an AI agent, how agents differ from chatbots and plain language models, the main types in use today and how to judge one.
AI Agent Statistics 2026: Live Usage, Growth and Activity Data
Current statistics on AI agents from public data: package downloads, editor installs, GitHub stars, pull requests opened by coding agents, release pace and benchmark results. Updated daily.
Best AI Coding Agents in 2026: Ranked by Usage, Benchmarks and Price
The AI coding agents developers use most, with live download and install numbers, benchmark results, entry prices and a way to choose between terminal, IDE and cloud agents.