PRIORITY AI-TOOLS Gateway What’s new Request an invite
עברית
Unified API for Priority ERP

Priority Gateway

Build the portal, the mobile app, the AI assistant, the nightly automation. One gateway and they all reach Priority the same way.

curl -X POST https://gateway/api/v1/query/form \
  -H "Content-Type: application/json" \
  -d '{
    "form": "ORDERS",
    "filter": { "STATDES": "Open" },
    "fields": ["ORDNAME","CUSTNAME","TOTAL"]
  }'

# → { "rows": [
#     { "ORDNAME": "SO20431", "CUSTNAME": "ACME", "TOTAL": 12450 },
#     ...
#   ] }

One gateway, four ways in. Read, write, run procedures, query across forms — over plain HTTPS.

Priority's current integration limits

Procedures, reports, actions, smart filters — WebSDK only

Running a procedure or a report means using Priority's own SDK. If your app is Python, .NET or mobile, it cannot get there at all.

No MCP layer for AI tools

Connecting Lovable, Base44 or any AI model to Priority needs a layer that Priority doesn't ship.

One gateway opens both.

How It Works

One gateway sits between your customer's apps and their Priority. Nothing to install on their side, no driver on anyone's laptop, and no language you are locked into.

Web / Mobile Apps n8n Workflows Dashboards AI / LLMs App Generator Claude.ai Priority Gateway REST API WebSDK SQLI MCP Speaks Priority on the inside. Speaks plain HTTPS to everything else. Hosted for you · nothing to install Priority ERP Forms · Procedures Reports · Documents Data · Business Logic

Four ways into Priority

One way in for your code, one for questions that span forms, one for AI tools. The same sign-in behind all three.

Read and write Priority from anything that speaks HTTPS

The portal, the mobile app, the n8n flow, a Python script: they all call the same address. Records, procedures, reports, documents, file uploads. The gateway works out how to fetch each one.

Typical callers
Customer portals Mobile apps n8n flows Python / .NET ETL jobs
POST /api/v1/query/form
{ "form": "ORDERS",
  "filter": { "STATDES": "Open" } }
REST API WebSDK OData

Ask questions that cross forms

Real SQL against Priority's own engine, with no driver on anyone's machine. It can only read: anything that would write is rejected before the call leaves the gateway. Dates, decimals and Hebrew come back readable rather than as raw Priority numbers.

Typical callers
Dashboards BI tools Analytics jobs Data exports LLM agents
POST /api/v1/sqli/query
SELECT ORDERS.ORDNAME, CUSTOMERS.CUSTDES
  FROM ORDERS, CUSTOMERS
 WHERE ORDERS.CUST = CUSTOMERS.CUST FORMAT;
SQLI Read-only

Talk to Priority in plain language

Connect Claude.ai or Claude Desktop, and your customer's IT manager can ask their own Priority a question and get a real answer. Or build them an assistant of their own on the same footing.

Typical callers
Claude Desktop claude.ai Lovable Base44 Custom AI agents
▸ You     Show open orders for ACME.

  Claude  priority_action(query, ORDERS,
            { CUSTNAME: "ACME" })
MCP Claude.ai

Sign in once — the app never holds a password

A server signs in once and gets back a token that expires on its own. A browser app opens our sign-in window instead: the person logs into Priority inside it, and the app receives a token you can revoke whenever you like. The password never reaches the page.

Typical callers
Customer portals Mobile apps Generated apps BI dashboards
POST /api/v1/auth/login
→ { "token": "gws_…",
    "expiresAt": "2026-08-08T18:20:00Z" }
Bearer sessions Connect popup

Three more capabilities

Added since the gateway's first release.

Ask business questions in plain language

Before the model writes a line of SQL, the gateway hands it a map of the part of Priority the question actually touches. So "how much did we sell last month" comes back right the first time, instead of plausibly wrong. Off by default.

Serve concurrent callers

Priority's SDK allows one login per process, so calls used to queue up behind each other. Switch on the worker pool and they run side by side instead.

The gateway teaches its callers

The gateway explains itself on request: which query dialect to use, which approach fits a given job, how to build an n8n workflow. A model looks that up mid-conversation instead of guessing.

A built-in n8n node for Priority integrations

Shopify orders into Priority. Nightly invoice exports out. Approvals routed through Slack. Drag the n8n-nodes-priority-gateway node onto the canvas and connect it. No glue code, and nothing to build per customer.

  • Everything you would do by hand. Read, create, update, delete and upload. Procedures, reports, form actions, and a query builder you click together.
  • The dropdowns fill themselves. Pick a form and its fields appear. Pick a field and its values load. Nothing to type by hand, nothing to keep in sync.
  • Subforms at any depth. Order lines, document lines, allocation rows: reachable however deep they sit.
  • The builder works out the joins. Pick a column from a related table and it finds the path itself. The gateway also hands an AI the node's own spec, so it can write you a working workflow without opening the editor.
  • It picks the fastest route for you. You wire the nodes; the gateway decides how to make each call. Override it when you want to.

v3.17.1 · 61 releases · MIT licensed

The node's parameter panel in n8n: operation OData, action Query, entity LOGPART — Part Catalogue, and an open field list showing Priority columns with their types.

Describe the app in one sentence. Hand it over the same afternoon.

You describe what the customer needs. It reads their forms, builds the app, tests it in a real browser, and hands back a zip you can deliver.

PHASE 1

Intake

Three questions settle the shape, the form and how people sign in.

PHASE 2

Discovery

It reads the real form out of the customer's Priority.

PHASE 3

Build

Builder agent scaffolds the app on disk under generated-apps/.

PHASE 4

Smoke test

It runs the app in a real browser and fixes what breaks.

PHASE 5

Handoff

Changes go back in at the right step, so nothing is rebuilt from scratch.

Four patterns × three targets

Pick the shape and the stack. It picks the template.

React HTML Python CLI
Web app forms & auth single-file —
Dashboard KPIs & filters read-only —
Mobile-first responsive PWA, offline-resilient —
Automation — — cron-friendly

Four subagents under the hood

The orchestrator dispatches one of these per phase.

AgentResponsibility
ResearcherPulls live form metadata from the Gateway
DesignerMaps forms and subforms to UI patterns
BuilderScaffolds files under generated-apps/
TesterWalks the scenario in real Chrome (≤3 retries)

Pre-built template library — growing every release

Reference apps the generator copies and customizes. New patterns ship in every release.

Dashboard React

orders-dashboard-react

Read-only orders dashboard with KPIs, search, and status filter.

Mobile-first HTML

stock-count-mobile-html

Mobile PWA stock count with subform writes and a finalize activation.

Automation Python CLI

invoice-export-python

A scheduled export of invoices to CSV.

Web app React

employee-order-react

Order entry for your own staff.

Web app React

auth-contact-react

A form your customers' own contacts sign in to.

Web app React

auth-priority-user-react

A starting point for Priority sign-in.

Web app React

contacts-list-session-react

A searchable list that keeps its place as you scroll.

+ more coming

New patterns ship in every release.

Connect AI to Priority ERP

Your customer wants AI that knows their Priority. Connect Claude.ai or Claude Desktop and they have it, in their own words.

1

Discover

The model learns your forms and fields as it needs them.

➔
2

Ask

"Show open orders for ACME from this month." Natural language in, structured rows out.

➔
3

Act

Create records, run procedures, produce reports and upload files, all in one conversation.

Works with

Claude.ai

Chat with live Priority data — ask, query, create.

Lovable

Generate web apps that read and write Priority through the gateway.

Base44

Internal dashboards on top of Priority — no backend code.

n8n + AI nodes

AI judgement inside a workflow: routing, alerts and triage.

Any MCP client connects. The gateway handles the sign-in and the routing. Full MCP details →

What partners are already shipping

Real customer asks already running on the gateway.

Order intake automation

Orders arrive from a webshop, EDI or email and land in Priority, checked and booked.

Mobile stock count on the floor

One HTML file on the warehouse phones, built from a single prompt. It survives a closed tab.

Ask the ERP in plain English

"Which orders shipped late this month?" Claude reads live Priority data and answers — no custom report.

Customer self-service portals

Orders, invoices, and delivery status — built in Lovable or Base44, served by the gateway.

Nightly invoice exports

A scheduled job pulls invoices to CSV and drops them on S3, SFTP or FileSmile. Built from one sentence.

Document generation on demand

Account ledgers and statements, triggered from an outside system over HTTP.

Where it runs

Hosted for you on Amazon Web Services with Cloudflare in front. There is no gateway for you to install, patch or keep running.

Hosted on AWS

The hosted gateway runs in Amazon's Tel Aviv region. Passwords and keys are kept in Amazon's encrypted store, never in a config file.

Cloudflare in front

Every request comes through Cloudflare first. It encrypts the traffic, absorbs attacks and filters out bots, and it shortens the trip from wherever your callers are.

No open front door

The servers accept no incoming connections at all. They dial out to Cloudflare, and traffic comes back down that same line, so there is no public door to find. Admin screens sit behind a second lock.

Running Priority on your own servers? That is fine — the gateway reaches it from here over an outbound connection you approve, so nothing has to be installed and no port has to be opened on your network.

Part of the Priority AI-Tools suite

Sign up once. After that your team reaches Priority from Claude, ChatGPT, Gemini, n8n or a live dashboard.