Loading content...
Internal Amazon platform that was on track to reshape how engineers understand codebases—live dependency graphs, CI awareness, and repo intelligence across ~2k repos, in the same spirit as OpenClaw-style tooling for the whole company.
a.amazon.dev was not a side dashboard—it was the beginning of a bet that internal Amazon development could feel as immediate and legible as the best open-source AI dev stacks. Think OpenClaw-class clarity—repo truth, dependency reality, and activity signals surfaced in one place—but wired to Amazon’s private git mesh, CI estate, and thousands of internal packages. The goal: make every engineer able to see the graph they were standing in, before they shipped or refactored.
At full stride the system continuously indexed ~2,000 repositories, ~900 CI/CD pipelines, and ~600 internal websites, turning that footprint into dependency graphs, churn signals, contributor maps, and git metrics you could actually act on—not quarterly reports, but day-one context.
The platform crawls Amazon's internal git infrastructure to identify and classify repositories by type (npm monorepos, frontends, Jupyter notebooks, GenAI/AgenticAI projects) and runs analysis jobs against each one. That is the mechanical layer; the product story is decision speed: teams could answer which repos depend on this internal package?, how hot is this codebase right now?, and who actually owns the blast radius? without spelunking through a dozen tools.
Job processing is handled by BullMQ workers backed by Redis, with results stored in DynamoDB. The frontend is a React SPA using tRPC for type-safe data fetching, Material UI with @emotion for styling, and React Query for cache management. Infrastructure is defined with AWS CDK and the monorepo is managed with NX.
CloudWatch RUM provides real-user performance metrics for tracking adoption and catching regressions—important when you are trying to earn the kind of default placement in an engineer’s workflow that tools like OpenClaw are chasing in the open ecosystem.
Locked behind Amazon's internal VPN — no public demo available. What remains on this page is the shape of a platform that was about to revolutionize how internal teams reasoned about code at scale: one URL, one graph, one place to stop guessing.
Working on this sparked a personal project. The experience of building repository analysis tooling inspired SaaS Recipes—a separate, independently built collection of curated SaaS templates and starter kits with detailed tech stack breakdowns.