Claude Code, our great ally in the SDLC for iMBranded

In visual communication and branding, where image is both the product and the engine of business growth, keeping a platform stable, up to date and aligned with user needs demands constant rigour.

Diverger built an iOS mobile application for iMBranded, with a Flutter frontend and a .NET backend, that fully digitalised the site survey process. The tool let sales staff record images and measurements of each location on site, place branding products over real photographs of the space to preview the result before producing it, and generate the PDF documentation automatically. In other words, a single in-house tool brought together data capture, the visual proposal and the commercial documentation.

The challenge we faced was, on the one hand, sustaining a product already in production, distributed through App Store Connect and deployed across three separate environments. And on the other, continuing to evolve it significantly while keeping the same number of developers in the squad —between four and six members, working under an Agile framework— and doing so without compromising quality or the delivery dates agreed with the client.

Faced with that challenge, we identified this as an ideal use case for bringing Claude Code into the developer squad as an ally across the whole software development lifecycle (SDLC). From reading the ticket defined in Jira through to deployment, taking in implementation, tests, code review and documentation. Our main goal was not to use Claude Code at one particular stage of the SDLC, but to agentize the entire workflow, end to end.

The working model we adopted drew heavily on the Anthropic ecosystem. Claude Code, the Claude API in CI, the Figma MCP server and a governance system versioned alongside the code, based on a CLAUDE.md of more than 120 lines acting as the source of context and persistent project rules, plus more than 7 architecture skills, custom slash commands and hooks.

We also created Ralph, an in-house automated code review agent built on the Claude API.

This meant that every time a developer proposed a change to the project (a pull request), Ralph reviewed it as a senior developer would, spotting errors, inconsistencies and departures from the defined architecture standards, and leaving its comments directly on the proposal. Ralph acted first in the developer's own local environment, before the change was submitted, so many problems were fixed there. And then automatically in the pipeline, as a mandatory validation step in the software build. No change could be merged into the project without passing its review first.

After several weeks of adoption, with Claude Code appearing in GitHub as co-author on more than 50% of commits, its impact translated into a test suite multiplied by nine against the starting point, an improvement of over 50% in code generation productivity, and test coverage above 80% in both backend and frontend. That allowed us to double the number of monthly releases with no significant incidents, with technical and functional documentation that was more complete, more current and delivered to the client in less time.

As Claude Code took on much of the implementation, testing and initial code review, the technical team could focus its effort on the tasks where its judgement added most value: understanding the client's problem, defining the technical requirements in detail, taking architectural decisions and reviewing code in depth, among others. The team shifted towards a role centred on the what and the why, rather than the how.

Adopting Claude Code in the SDLC for iMBranded therefore let us demonstrate that it is possible to increase the pace and volume of delivery without giving up quality or the judgement of the engineers. By integrating a set of agents at every phase of development and governing it through rules versioned alongside the code, we turned each developer's individual judgement into a shared project standard, verifiable at every step.

This success story has not only held up with iMBranded to this day: it has proved replicable, carrying over to other teams and technology stacks where agile, continuous delivery is required without sacrificing quality or commitment to the client.