The project's README from github.com/emoss08/Trenova. As one markdown file: README.md.
Source-available trucking TMS written in Go, PostgreSQL, GraphQL, and React.
Dispatch, rating, billing, accounting, EDI, FMCSA compliance, and document automation for asset-based carriers.
Website · Features · Pricing · Self-hosting · Discussions · Discord
[!IMPORTANT] Trenova is pre-release software under active development. It is not yet suitable for production use. Follow releases or join the waitlist for updates.
Trenova is a transportation management system (TMS) for small and mid-sized over-the-road carriers in the United States. Enterprise TMS platforms often cost more than $50,000 a year, which puts them out of reach for the carriers that make up most of the industry. Trenova aims to close that gap with a flat-priced, self-hostable TMS that covers the full order-to-cash cycle: shipment entry, dispatch, driver assignment, rating, billing, settlement, accounting, and compliance.
The source is public under the Fair Source Functional Source License, so carriers and integrators can read the code, audit it, and run it on their own infrastructure.
See the full capability list at trenova.app/features.
| Layer | Technology |
|---|---|
| API | Go 1.26, Gin, gqlgen GraphQL |
| Database | PostgreSQL via pgx, SQL-first migrations |
| Search and cache | Meilisearch, Redis (JSON and Streams) |
| Change data capture | GTC: a Go PostgreSQL logical-replication connector that projects rows into Meilisearch and Redis (runs as the ghcr.io/emoss08/gtc image; routing in deploy/gtc/sinks.yaml) |
| Web client | React 19, TypeScript, Vite, GraphQL |
| Documents | Gotenberg (PDF), MinIO / S3-compatible object storage |
| Deployment | Docker images on GHCR, Caddy, Docker Compose |
services/tms Core TMS API (Go)
services/samsara-sim Samsara telematics simulator for local development
services/edi-partner-sim EDI trading-partner simulator
shared/ Shared Go packages (money, geo, dispatch planner, PC*MILER, AS2, ...)
client/apps/web Web application (React)
client/apps/dash Admin dashboard (React)
deploy/ Dockerfiles, Caddyfile, observability stack
docs/ Engineering guides referenced by CLAUDE.md, Bun notes, product guide
Prerequisites: Go 1.26+, Node.js with pnpm, Docker, and Task.
git clone https://github.com/emoss08/trenova.git
cd trenova
cp .env.example .env
# Start PostgreSQL, Redis, MinIO, Meilisearch, and Gotenberg
docker compose -f docker-compose-local.yml up -d
# Run the TMS API
task tms:run
# Run the web client
cd client && pnpm install && pnpm dev
If a build uses all your RAM and CPU, turn on low-resource mode. Go then compiles one package at
a time on one thread and runs its garbage collector harder, golangci-lint and the test runners
are capped the same way, hot reload waits 2 s after the last save, and docker-up starts only
PostgreSQL, Redis, MinIO and Temporal. Builds take longer, but the machine stays usable.
export LOW_RESOURCE=true # every task from now on (add it to your shell profile)
task tms:run-watch LOW_RESOURCE=true # or a single command
task tms:docker-up-minimal # just the minimal containers, in any mode
On Windows, setx LOW_RESOURCE true sets it for every new terminal.
LOW_RESOURCE_JOBS=2 allows two packages at a time with two threads each; that is a sensible
setting on a 16 GB machine. Run only the client app you are working on
(pnpm --filter @trenova/web dev) rather than pnpm dev, which starts both.
Run task list to see every available task. Agent-facing engineering guides live in docs/engineering. Operations runbooks, design notes, roadmaps and research live in trenova-documentation.
Container images are published to GitHub Container Registry on every release:
ghcr.io/emoss08/trenova/tmsghcr.io/emoss08/trenova/clientDeployment files (Dockerfiles, Caddyfile, observability) are in deploy/. See trenova.app/self-hosted for the self-hosting overview. Trenova Cloud production servers use a separate private deployment repository that pins released image tags.
Trenova is Fair Source for transparency and community review, but external code contributions are not accepted at this time so the core team can keep the architecture and roadmap consistent. You can still help by:
See CONTRIBUTING.md for details.
Trenova is licensed under the Functional Source License, Version 1.1, Apache 2.0 Future License (FSL-1.1-ALv2). You may use, self-host, and modify Trenova, including in commercial environments. You may not offer it as a hosted service or use it to build a competing product. Each release converts to the Apache 2.0 license two years after publication.