Business Central mobile flows

How it maps to Plur-e Mobile

Plur-e Mobile talks to Business Central through one Plure API endpoint (erp/processrequest) and one AL bound action, passing a ProcessMethod name and a Parameters array. The MCP server does exactly the same, with the same identity the app would use: the caller's Admin Center account is mapped to a Plur-e Mobile user by email, and its WMS / Sales module assignments provide the WarehouseEmployee and SalesPerson that Business Central expects. Tenants without mobile users set defaults under AI & MCP → Settings → ERP identity.

Every section tool takes the same arguments:

{
  "name": "bc_wms_receipts",
  "arguments": {
    "method": "GetWarehouseReceipt",
    "parameters": { "WsheDocumentNo": "WHSE REC-000123" }
  }
}

parameters is an object (sent as a one-element array, as the app does) or an array when the skill says the array is the payload (count lines, pick lines, journal lines). The tool validates that the method belongs to the section, injects the identity keys when the method needs them, and returns Business Central's answer as JSON. Base64 PDFs and Excel files come back as an embedded resource; lists longer than 500 rows are truncated with truncated and totalCount.

Skills

The plugin ships nothing: every skill lives on the server and is gated by the tenant's licence and the web roles allowed on each skill (AI & MCP → Skills).

Three ways to read them: the bc_get_skill tool ({ "name": "sections/wms-receipts" }), the MCP resource plure-bc://skills/sections/wms-receipts, or the prompt skill_bc_wms_receipts. bc_list_methods lists the methods of a section or searches the catalogue by text.

Editing skills in the Admin Center

AI & MCP → Skills lists every built-in document of both servers (Business Central and CRM). A tenant administrator can customize one (the built-in text is kept as revision 0 and every save adds a revision, so a reset is always possible), restrict it to web roles, disable it, export it as Markdown, import a .md file, or add custom skills of the tenant (flows/…, sections/… or a guide at the root). The MCP server serves the effective document — tenant copy, then a Plur-e global override, then the build — through bc_get_skill, the resources and the prompts, within about 15 seconds of a change. A tenant's AI agent can further narrow which of the 16 sections a connection sees, on top of what the user's MCP role already allows.

Scanning: codes from an image

Labels on a shelf, a pallet or a delivery note can be read from a photo instead of one scan at a time.

• bc_decode_barcodes({ image: { base64 | url | handoffId }, hints? }) finds every barcode in the picture (EAN/UPC, Code 128, GS1-128, QR, Data Matrix, PDF417…), keeps their position and returns them in shelf order (top-to-bottom, left-to-right) with the GS1 application identifiers already split (01 GTIN, 10 lot, 21 serial, 00 SSCC, 17 expiry). URLs are limited to the tenant's allow-list (SharePoint, Azure Blob, plur-e.com by default) and images to 4 MB.
• bc_resolve_codes({ codes, locationCode? }) turns each string into a typed reference — bin, item, serial, licensePlate, document (receipt, pick, put-away, transfer, sales or purchase order by prefix) or unknown — by asking Business Central the same way the app does (GetBinByLocation, IsItemOrSerial, GetItemIdentifier, GetLicencesPlate).

A typical count: photo of the shelf → bc_decode_barcodes → bc_resolve_codes → the bin and the items are known → bc_wms_inventory_counts writes the lines. The flows/count-bins-from-photo skill walks through it.

Handoff: camera, signature, photo, file without login

When the user is at the shelf or with the customer, the server opens a one-time page of Plur-e Web (https://app.plur-e.com/h/<token>): valid 15 minutes, single use, no login, bound to the caller's tenant and user. Nothing from Business Central is shown on it.

Clients that support URL elicitation open the page and the tool waits for Done; other clients receive the link and call bc_handoff_result(handoffId) later. Assets are stored in the tenant's database for 24 hours and only the MCP server can read them: in any FileInfo array, { "HandoffId": "…", "AssetId": "…" } replaces the base64 and the server fills it in.

Roles and audit

• Section tools are declared write, but each section's area starts at Read only in a role: a user whose MCP roles grant no Write on that area can call read methods only and gets scope_denied immediately on writes.
• Granting Write on an area in a role enables that section's non-posting write methods for whoever has the role; granting Post additionally enables the methods marked posting in the skills (post receipt, register pick, post count, post journal, ship order), which cannot be undone in Business Central. There is no approval queue: a write runs the moment a role allows it.
• Every call is audited with the method and a summary of the parameters (base64 payloads are stripped) under AI & MCP → Audit log.
• Errors carry a machine-readable code: invalid_method (with a suggestion), invalid_arguments, erp_user_not_mapped, bc_error (Business Central's message), handoff_pending, handoff_expired.

Printing labels

bc_list_printers returns the tenant's printers (Plur-e print queue) and the templates with the fields each one needs; bc_print_document(printer, template, documentNo, fields) queues the label — LP labels (ZPL LPS), item barcodes (BarCode) and box / address labels — exactly as Plur-e Mobile does, and the Plur-e Print agent on the workstation prints it. It is a write tool: it needs Write on the documents area. Packing lists, receipts and vouchers are PDF renderings that only Plur-e Web and Mobile produce; the tool refuses those templates.

Not covered

Offline vouchers, package shipments through carriers and the Stripe payment screen of Plur-e Mobile do not touch Business Central and are not exposed.

The server's tools/list and the index skill are the authoritative catalogue; counts above come from the generated _catalog.json of the server build.

Related: Business Central tools · MCP roles · Connect your AI client · Security model