ollie

Overview

Ollie is a self-hosted freight Transportation Management System — the operational core of a trucking dispatch desk: loads, trips, drivers, trucks, trailers and facilities, in a single Rust service you run yourself.

The thing that makes it different is that it’s MCP-first. Most software bolts an AI assistant on at the end; Ollie was built so an agent can run the back office through a Model Context Protocol server as a first-class surface, with the same auth and the same data the humans use. Dispatch a load, answer a question over the fleet, ask a question about a scanned document — all of it is callable by a tool-using agent.

How it works

Three API surfaces, one domain. Pick the caller:

• Fleet MCP (/fleet/mcp) — for AI agents and tool-using assistants. The preferred surface.
• Fleet REST (/fleet/api/v1) — for the Fleet web app and programmatic integrations.
• Driver portal (/driver/api/v1) — for the driver mobile PWA only, with passkey or PIN auth.

The domain. Loads move through a lifecycle (planned → assigned → dispatched → in_transit → delivered → invoiced → settled) with stops, rate line items and detention tracking. Trips are the driver-facing execution of those loads — stop arrive/depart, check-calls, relay chaining, and deadhead/loaded mileage via OpenRouteService HGV routing. Facilities are geocoded and de-duplicated. An append-only event journal records what happened, and a set of doctors diagnose-and-repair data integrity when something drifts.

Documents that answer questions. Rate cons, BOLs, PODs and photos are stored content-addressed and deduplicated on disk, then processed in the background by Ollama: text summaries and vector embeddings for semantic search, with images and scanned PDFs routed through a vision model. Vector search runs on LanceDB, so you can ask natural-language questions over a document.

Stack: Rust · Axum · LanceDB · Ollama · rmcp (MCP) · AGPL-3.0.

The running log below is the week-by-week build history.