How We Built Charlie, Part 15: What We Learned Building Charlie
What we learned about making AI engineering work owned, bounded, verifiable, and useful to a team.
How Coforma uses Charlie across daemons, PR reviews, and dozens of application and infrastructure repositories.
Coforma connects private-sector technical experience with public-service missions, helping agencies modernize the systems behind essential services. Its human-centered approach makes those services clearer, more inclusive, and easier to use at scale.
Coforma uses many AI tools and agents for different use cases, with Charlie working in the shared spaces and context where human engineers work. The team describes how daemons, PR reviews, and multi-repo support have become part of its engineering system.
Coforma frames that role as a shift from individually prompted tools to autonomous AI engineers working with the team’s shared context.
Charlie is one of the pioneers of the second wave of AI tools. The first wave was individual assistants, chatbots, and coding tools that require continual prompting and attention to move work forward. They’re useful, but have a major observability gap, especially for enterprise engineering applications. The second wave is autonomous AI engineers that work in the shared spaces and context where human engineers work, and are not restricted to individual laptops and local environments. This makes them superior for scalable agentic engineering on teams that ship production-ready code in compliant and secure environments.
– Coforma
The team quantifies Charlie’s place in its architecture through eight months of contributions.
In the last eight months Charlie has opened 1,538 pull requests and merged 1,198, done 9,264 PR reviews, and made 3,561 commits, totaling over 15,000 contributions to our codebase and forming a critical part of our architecture. The three features that are most valuable to us are: the daemons, PR reviews, and multi-repo support.
– Coforma
For Coforma, daemons handle recurring engineering work and security research in the background.
The daemons run continuously in the background doing chores like keeping our documentation updated, triaging incoming Linear issues and bug reports, keeping pull requests mergeable, and responding to engineering alerts in Slack. We also built an autonomous security researcher that does a daily sweep of security blogs, scans our repos for signs of vulnerabilities, and leaves us with remediations ready to go—saving us human time and keeping our codebase on the cutting edge of security.
– Coforma
The team values PR reviews that are relevant, high-quality, and aligned to its engineering standards.
Charlie’s PR reviews are better than any other AI reviews we’ve used. They’re consistently high-quality and more relevant than those from the other AI tools we’ve tried. And even when other bots are leaving code reviews, Charlie keeps them aligned to our engineering standards by double-checking their work. Our team merges PRs with more confidence knowing Charlie daemons have reviewed and approved our code.
– Coforma
Coforma ties that relevance to Charlie’s context across dozens of application and infrastructure repos.
The multi-repo support is one of the reasons Charlie-authored code and reviews are consistently more relevant and production-ready. Unlike other AI tools that limit the context of sessions to a small set of repos, Charlie always has the context of our dozens of repos and is able to make judgments about which repos relate to the current task. It can take a brief feature request and create an integrated changeset across many application and infrastructure repos, even when the human doesn’t know which repos to use.
– Coforma
Coforma’s description of Charlie ends with its role as a foundational agent engineer.
Like most engineering teams, we use many AI tools and agents for different use cases. Charlie is the foundational agent engineer that autonomously gates everything and keeps all contributions (whether by humans or AI) aligned to our standards, our compliance obligations, and our business strategy.
– Coforma
Coforma’s account describes an engineering system where people and many AI tools contribute in different ways, while Charlie keeps the work aligned across repos, reviews, and recurring tasks. That shared context is what makes Charlie a foundational part of the team’s architecture.