The MCP server
for logistics AI
Give AI agents deep context about your shipments, carriers, tenders, and rates — without your data ever leaving your infrastructure.
# Install and initialize
$ npm install -g atlas-mcp && atlas init
# Or run via Docker
$ docker run -d \
-v ./config.yml:/app/config.yml \
cargofy/atlas:latest
// ATLAS MCP server running on :3000
// 36 tools available
// 21 logistics data models loaded
// 2 connectors active, syncing every 15m
36
MCP Tools
21
Data Models
Apache 2.0
License
0 bytes
Leave your infra
Why ATLAS
Enterprise logistics won't share their data. They're right.
Carrier relationships, pricing history, client contracts — that's competitive advantage. No compliance team signs off on sending it to a SaaS vendor.
Before — Without ATLAS
AI agents start every conversation from zero. No carrier history, no rate context, no operational memory. Your team rebuilds context manually every time.
After — With ATLAS
ATLAS runs inside your security perimeter. Indexes your TMS, contracts, and rate sheets locally. AI agents query your own data through a standard protocol. Nothing leaves your network.
Architecture
Your data. Your infrastructure. Standard protocol.
ATLAS answers questions. Your AI platform acts. The boundary is clean.
├── TMS / WMS / ERP
├── Contracts & Documents ←── MCP Protocol ──→ Claude / GPT / Custom
├── Rate sheets & history
└── ATLAS v1.0
├── Connectors (REST, filesystem, SSE)
├── 21 models · incremental sync
├── SQLite (dev) / PostgreSQL (prod)
└── 36 MCP tools · /api/health · webhooks
Data Models
8 core models. 13 domain extensions.
Schemas grounded in project44, DCSA, UN/EDIFACT, GS1 EPCIS, and IATA. Enable only what you need via config.
Core — always available
// Shipment
Freight movements end-to-end
// Carrier
Carriers + rating + capacity
// Lane
Origin–destination pairs with history
// Rate
Freight pricing and benchmarks
// Document
BOL, CMR, invoice, customs
// TrackingEvent
GPS + status events + exceptions
// Party
Shipper, 3PL, consignee, broker
// ServiceLevel
SLAs, transit time, KPIs
Extensions — opt-in via config.yml
MCP Interface
36 tools. Every logistics context your agent needs.
Organized in 8 groups: discovery, shipments, carriers, SLA monitoring, procurement, marketplace, post-shipment, and generic query.
Discovery
Shipments
Carriers & Rates
SLA & Ops
Procurement
Marketplace
Post-shipment
Generic
Platform
Production-ready out of the box.
Not a toy. Built for operators running real freight every day.
SQLite + PostgreSQL
Default SQLite, zero config. One config line to switch to Postgres for production.
Webhooks
Push SLA violations and issues to any URL. HMAC-signed. Automatic retry on failure.
Scoped Tokens
Bearer tokens with per-scope permissions. Read-only agents, write agents — your call.
Health Endpoint
GET /api/health — per-connector status, record counts, uptime. Drop into any monitor.
Incremental Sync
Connectors track last_synced_at and only fetch new records. No full scans.
Composite Mapping
One API response → multiple entities. Shipment with nested scan events in one pull.
Connectors
Connect your existing systems.
Incremental sync, composite mapping, exponential backoff retry. No data migration required.
CLI
First-class CLI. Zero config to start.
From install to running MCP server in under 60 seconds.
atlas initInteractive setup wizardatlas validateCheck config.yml for errorsatlas seed ./data/Load JSON / CSV / XLSXatlas statusRecord counts + sync timesatlas startStart HTTP + MCP serverATLAS Status
shipments 4 823 records
carriers 241 records
tracking_events31 402 records
tenders 18 records
rates 1 097 records
last sync: 3 minutes ago
Open standard.
Built for the industry.
Apache 2.0. Self-hosted. Contributions welcome.
Built by Cargofy, open for every logistics company.