// the find
suxrobGM/logistics-app
AI-powered fleet management platform with an agentic dispatcher that matches loads to trucks, ensures HOS compliance, and optimizes routes. Built for trucking companies.
LogisticsX is a full trucking TMS — dispatch, invoicing, payroll, compliance — built on .NET 10 with CQRS/DDD, topped with an AI dispatch agent and an MCP server. It's for small-to-mid carriers replacing spreadsheets and disconnected tools, and for .NET developers wanting to see a real multi-tenant SaaS architecture; it's not much use to Angular or mobile developers, since those clients are closed-source.
The backend is genuinely substantial for something built mostly solo: 372 endpoints, CQRS with 212 commands and 155 queries, 1000+ automated tests, and architecture tests in the test/ folder to keep layering honest. Per-tenant database isolation is a sound call for a product handling payroll and invoicing — no shared-schema query can leak another tenant's financial data. The dispatch agent requires human approval before committing an assignment and the docs describe a 31-tool agent surface, which suggests someone thought about the actual failure mode (a bad truck-to-load match costs real money) instead of just wiring an LLM to the database. Documentation is unusually thorough and looks maintained alongside the code — architecture docs, API reference, per-integration deployment guides, and active build/deploy CI.
"Open source" is doing a lot of work here: the license is PolyForm Noncommercial, and the actual product UI — the Angular TMS/customer/admin portals and the Kotlin driver app — sits in a private submodule you don't get. What's public is the backend only; you can read the domain model but can't run the thing most people picture when they hear TMS. One Postgres database per tenant scales cleanly in isolation terms but linearly in operational cost — migrations, backups, and connection pools all multiply with tenant count, fine at 10 tenants, a real tax past a few hundred. The project leans hard on its AI story (dispatcher, copilot, MCP server, three LLM providers) but its own roadmap docs list cost guardrails and an eval harness for the agents as not yet built, meaning the safety net around those autonomous LLM calls isn't there yet, just the plumbing. The topic list includes `maui`, but the README says the driver app is Kotlin Multiplatform/Compose — a small sign the repo metadata hasn't kept up with an architecture that's clearly shifted more than once.