Orbit · Platform

One platform. Four layers.
Every app shares it.

Orbit isn't a bundle of apps — it's a unified platform. The same AI engine, the same integration backbone, the same JWT-scoped security powers BPI, MFT, AI-OCR, and Vault. Build once, govern once.

4
apps · one platform
6
real AI surfaces
9
protocols · every major format

Architecture

A layered platform, by design.

Four horizontal layers. Each one earns the layer above it. AI shouldn't be glued on top — it should run through every flow, every parse, every mapping. So that's how Orbit is built.

The Orbit platform architecture: four stacked layers — Applications (ORBIT BPI, SecureFlow MFT, Orbit AI-OCR, Orbit Vault), AI Engine (MAPR AI, JSONata Assistant, AI-OCR Engine, Doc Insights), Integration (formats, protocols, process flows, dynamic endpoints), and Security & Identity (JWT/SSO, tenant isolation, RBAC, audit log).
01

Applications

Four independent apps that run on the same Orbit backbone — subscribe to one or combine into a full suite.

ORBIT BPISecureFlow MFTOrbit AI-OCROrbit Vault
02

AI Engine

Real, named AI surfaces that ship inside the platform — not a buzzword roster.

MAPR AIJSONata AssistantAI-OCR EngineDoc Insights
03

Integration

Every format we parse, every protocol we speak, every flow we run — same engine.

X12 / EDIFACT / IDocAS2 / SFTP / OFTP2 / RESTProcess FlowsDynamic Endpoints
04

Security & Identity

JWT-scoped, tenant-isolated, RBAC-governed, audit-logged — from day one.

JWT / SSOTenant IsolationRBAC ProfilesAudit Log

AI surfaces

Real AI. Named. Shipped.

Other platforms list "AI-powered" as a bullet. Orbit ships six specific AI surfaces that have names, jobs, and real impact in production today — built on the Vercel AI SDK with Claude / GPT-4 / Gemini, depending on the workload.

MAPR AI alone replaces an entire integration consultant — a claim none of our competitors will make about their AI.

Four AI surfaces inside Orbit: MAPR AI mapping X12 850 to IDoc ORDERS05, the JSONata Assistant turning a sentence into a JSONata expression, the AI-OCR engine extracting a typed message type from an invoice PDF, and Doc Insights rendering an EDI transaction as a business-friendly view.

MAPR AI

The mapping agent

Hand it two sample payloads. It parses both schemas, proposes a complete rule set, dry-runs it, repairs failures, and applies a live mapping in your IDE. 80 hours of work in 8 minutes.

JSONata Assistant

Expression copilot

Describe the transformation in English. The assistant returns valid JSONata you can paste into a rule — fallbacks, qualifiers, conditionals, all of it.

AI-OCR Engine

PDFs become typed data

Upload a document, chat to refine the extraction schema, save it as a message type. Runs at flow time so production PDFs become structured data with the same shape every run.

Doc Insights

Business-friendly view of EDI

Renders raw X12 / EDIFACT / IDoc as a clean, business-readable card stack. Buyers, finance, ops can read a transaction without reading EDI.

Dashboard Agent

Chat over your live data

Every BPI dashboard surface has an Orbit-grounded assistant: "how many failed transactions today", "recent executions for ACME", "connection health" — answered against the live tenant data.

Cron Assistant

English in, cron out

Describe a schedule ("every weekday at 9am EST"). Get a valid cron expression with timezone. No more cron-tab Googling.

Security model

Multi-tenant from day one.

No "we'll harden it for enterprise" promises. The platform is tenant-scoped at the JWT level, RBAC-governed per-app, and audit-logged on every action. From the first request.

Diagram of Orbit's multi-tenant security model: two users in different tenants each present a JWT scoped to their tenant ID, route through the JWT Identity Gateway, and the gateway routes them only to their own tenant's data.

Tenant ID from the JWT, never from a header

Every request's tenant scope comes from a validated JWT claim. Headers can't escalate. The same rule applies to every Java backend (BPI / MFT / TPM) and the Bun runtime.

RBAC per app, with profile-level bundling

Roles bundle permissions. Profiles bundle roles, scoped to specific apps. Users get profiles, never raw permissions. The UI is gated by , the API by Spring Security.

Audit log on every action

Every meaningful action — login, role change, mapping activation, share link create, partner edit — lands in an audit_log table with actor, tenant, request ID, and outcome. Exportable, queryable, retained.

Integration backbone

Every format. Every protocol. One backbone.

X12, EDIFACT, IDoc, XML, CSV, JSON for data. AS2, SFTP, FTP/FTPS, OFTP2, REST, JDBC for transport. Plus 200+ pre-built connectors to SAP, Oracle, NetSuite, Salesforce, Shopify, Amazon, and the long tail of trading partners.

Two rows of icon cards: top row showing the formats Orbit parses (X12, EDIFACT, IDoc, XML, CSV, JSON) and bottom row showing the protocols Orbit speaks (AS2, SFTP, FTP/FTPS, OFTP2, REST API, JDBC), each with a +custom slot at the right.

Powering enterprise integrations worldwide

AmazonWalmartFordUSPSUPSDHLBest BuyCVS HealthComcastCarrefourJohn LewisASDADaimler TruckW.W. GraingerUnion PacificAsusACE HardwarePublixEDEKADB SchenkerRhenusFNACBUNZLUlineAmazonWalmartFordUSPSUPSDHLBest BuyCVS HealthComcastCarrefourJohn LewisASDA

See the platform.
Then run it.

Book a 30-minute walkthrough of the architecture, the AI surfaces, and the security model — with an engineer, not a salesperson.

Live MAPR AI demo
Real tenant-isolation walkthrough
Connector inventory review
No sales pitch
30 minutes flat
Engineer-led