Money Forward, one of Japan's largest cloud-based financial software companies, serves more than 17 million individual users and 400,000 businesses with products spanning personal finance, accounting, payroll, and invoicing. Founded in 2012 with the mission to "move your life forward," the company is evolving from digital transformation to AI transformation, with the goal of becoming Japan's leading back-office AI company.
Money Forward's engineering team had experience with AI coding tools, but encountered limitations. The challenge went deeper than IDE compatibility. When the team needed to migrate code, existing tools hit a wall because they only understood single projects within one IDE. Claude Code, by contrast, understood and navigated the full context across languages and repositories.
Money Forward launched MEPAR—Money Forward Engineering Productivity AI Research—with a clear mission: ensure every engineer gets a productivity boost.
Money Forward built a disciplined evaluation process within MEPAR, running trials with small groups of engineers and only rolling out tools that received strong feedback. For each trial, the team created a public channel where members could exchange thoughts and share how they were using the tools.
The team built an internal template to integrate Claude Code into PR workflows based on their style guide, then tested several options side by side. Teams consistently found Claude Code's reviews to be the most useful. Developers could fix issues before human review, which elevated overall PR quality and made reviewers' jobs significantly easier.
The trial's visibility across the company meant any employee could follow along. When trial members shared their results, more than 200 engineers from teams outside the trial requested access. The adoption became bottom-up instead of top-down—engineers pulled Claude Code into their workflows because they saw their peers getting real productivity gains.
After rolling out organization-wide in July and hosting a training workshop with Anthropic in August, Claude Code became essential infrastructure. The original trial channel became the official Claude Code channel and the most active AI-related channel in the company.
More than 80% of engineers have adopted Claude Code, with over 70% using it daily. The code acceptance rate exceeds 90%—engineers trust Claude Code's output enough to integrate it without extensive review. Engineers report they are saving on average 7 hours per week per engineer, according to an internal survey.
Claude Code now serves as the foundation for Money Forward's engineering productivity across complex project development, automation tooling, and developer experience. Engineers use it for full feature implementations, large-scale refactoring, PR review automation, and an automated onboarding system that handles frontend, backend, and end-to-end repository setup with a simple /onboarding command.
The CLI-based architecture proved essential for Money Forward's diverse IDE ecosystem. Engineers can use Claude Code with WebStorm, JetBrains IDEs, RubyMine, and VS Code without changing their preferred tools.
For the Public API Initiative, API endpoint implementation dropped from 2 days to 5 hours. Claude Code generated approximately 80% of the implementation code. Developers shifted from writing code to directing it—planning architecture, reviewing output, and making the judgment calls that matter.
The transformation in onboarding proved equally significant. Cloud Accounting Plus, one of Money Forward's fastest-growing products, expanded its development team from Tokyo and Kansai to include teams in India and Vietnam. Previously, onboarding meant hours of troubleshooting and senior engineers getting pulled away from their work. Now, a simple /onboarding command guides new members through the process interactively.
"Claude Code's agentic capability transformed onboarding from frustration to enjoyment," said Aaron Li, Staff Engineer. "What used to take a week now takes a day. And we've saved countless hours for existing team members who used to get pulled away from their other work to help someone get their environment running. From a developer experience perspective, the difference is profound."
Another Staff Software Engineer, Tamai Ayumi, built Money Forward's entire MCP Server—solo. “I built Money Forward's entire MCP Server in three months using Claude Code. It implemented my plans and reviewed code—essentially acting as my only colleague on the project. I couldn't have done it without Claude Code."
The team also built two internal tools: Sherlock, which helps investigate incidents by searching logs and surfacing past cases, and Lumos, an AI agent that automatically fixes Kubernetes issues and submits a PR within five minutes.
Money Forward plans to extend Claude Code across the full software development lifecycle—from product discovery and test generation to SRE operations and multilingual code explanations for distributed teams in Japan, India, and Vietnam.
"Speed is our core value and moat," said Tran Ba Vinh Son, Group Company CTO and Manager of MEPAR. "Developer productivity is therefore a business strategy: the faster we turn ideas into reliable products, the stronger our competitive position. Claude Code compresses cycle times, elevates code quality, and lets small teams ship bigger features with confidence. Our vision is an AI-assisted development model where Claude is an integral teammate across planning, implementation, testing, and operations. That's how we scale innovation across Money Forward."