Water Monkey turns cloud-cost investigation into a reviewable workflow rather than a list of opaque recommendations. The public project combines a React interface with a Rust/Tauri desktop backend, native keychain storage, persisted finding history, and optional local AI summaries through Ollama.
The problem
Small teams can accumulate unattached disks, reserved IP addresses, inactive projects, and unexpected API spend across several providers. Each platform exposes different inventory and pricing concepts, so identifying waste becomes repetitive and estimates can be difficult to trust.
My ownership
I own the product direction, interface, provider integrations, desktop backend, local persistence, test coverage, security boundaries, release preparation, and documentation. The project remains private while I prepare it for a future public release.
System design
The application uses React for the dashboard and Tauri with Rust for native desktop capabilities. Provider scanners cover AWS, GCP, Azure, Vercel, Supabase, and OpenAI. Credentials are stored in the operating system keychain rather than repository files, while scan settings, findings, and activity remain local to the application.
Findings persist across scans with open, acknowledged, resolved, missing, and reopened states. Provider scans report status independently, so one failing integration does not erase successful results or block a targeted retry.
Trust and safety decisions
Every cost estimate carries its pricing assumptions, and severity is derived from one documented rule. Connection tests are read-only. Destructive remediation is limited to explicitly supported resources and requires confirmation; review-only providers direct users to the appropriate platform instead of implying an unsafe level of automation.
Optional summaries run through a user-configured local Ollama endpoint. This keeps AI assistance opt-in and avoids making a hosted model a requirement for the core scanning workflow.
Release boundary
The working application remains in a private repository while development and release preparation continue. Any future public release will define its licensing, contribution surface, documentation, and commercial boundaries explicitly rather than implying that unfinished private work is already available.
Validation and operations
The project includes Vitest coverage for scanners, persistence, severity, credential behavior, deletion safeguards, and interface workflows, plus Rust tests and checks for the Tauri backend. Release documentation covers signed builds and automatic updates, while provider documentation records required permissions, pricing assumptions, and recovery steps.
Outcome and next iteration
Water Monkey currently provides multi-provider scanning, explainable estimates, lifecycle-aware findings, local activity history, guarded remediation, native credential storage, and optional local AI insight. It remains under active development; the next iterations focus on expanding safely supported resources, strengthening release automation, and refining decisions from real-world provider behavior.