Agents Index

Spark Inventory

OnlineNew

AI inventory management: stock, Demand ESP reorder recommendations and purchase orders.

Visit publisherConnectJSON

What it can do: 80 tools

  • draft_ruleDraft an automation rule from a plain-language description (e.g. 'block POs over $10k when created'). On success returns a DISABLED draft rule alongside any detected conflicts, an approver hint and the Spark bot it was f
  • list_standard_reports read-onlyThe standard reports run_standard_report can produce (the Reports page's ready-made answers): purchasing plan, inventory health, inventory position and runway, sales and margin, backorder risk, supplier reliability, inve
  • delete_manufacturing_orderDelete a manufacturing order, named by id (GUID) or number, as the manufacturing order screen does. Only a draft or canceled order that no work order and no inventory movement references can be deleted; cancel an issued
  • list_demand_evidence read-onlyList the tenant's demand evidence records — durable research artifacts that give demand signals provenance (a URL, an excerpt, a structured claim about a market condition). Each row carries the scope (Item/Tag/Category),
  • list_saved_reports read-onlyThe reports the signed-in user saved on the Reports page, newest first. Saved reports belong to one person, so this works only in the in-app assistant. Run one with run_saved_report.
  • list_edi_item_mappings read-onlyList the EDI item mappings (cross-references): which Spark item each trading partner's SKU or UPC stands for, with the customer's description, unit, the conversion to the Spark unit and the customer price. Filter by trad
  • start_integration_connectGenerate the provider authorization URL that starts connecting an integration (providers: shopify — requires shopUrl, amazon, qbo, xero, zoho_books, etsy). IMPORTANT: you cannot complete the connection yourself — hand th
  • import_customer_returnsRecord customer returns that were already processed in another system (a warehouse's daily returns feed, or a returns history) so demand velocity, daily consumption and the forecast net them. Without returns, demand is g
  • commit_labor_planCommit a run's scheduled labor draft as a reservation, as the labor planner screen's commit does: the crew it names can then start work on it from their my-work page, and its hours count against everyone else's plans. Re
  • move_stockMove stock between locations — bin to bin inside a warehouse, or straight into another managed warehouse — the Inventory page's Move action. Only available (unreserved) units move. Each line names the item, quantity, sou
  • get_labor_work_day read-onlyOne worker's day, as their my-work page shows it: the runs they are committed to that day with their part of each crew, the work session running now, time recorded that day, and their open issues. day is a calendar date
  • create_supplier_price_bookCreate a vendor price book (rate card): name, optional per-tenant-unique code, currency, active window, interpolation mode, supported run quantities (runQuantityLimits), and either the spec attribute schema that keys its
  • list_supplier_price_books read-onlyList the org's vendor price books (rate cards): id, vendor, client attribution (null = shared house book), currency, active window, supported run quantities (runQuantityLimits), and the spec attribute schema. Archived (s
  • cancel_purchase_orderCancel a purchase order that is Open, Created, Waiting Approval or Issued and has no delivery receipt (an order already being received is handled in receiving). The preview shows the order, vendor, total and status chang
  • close_out_campaignRecord a finished campaign's actuals: spend (with currency), impressions, clicks and attributed orders — at least one. Close-out opens the day after the campaign's last day in the workspace's time zone (see campaign.clos
  • create_purchase_orders_from_recommendationsConvert reorder recommendations into draft PURCHASE ORDERS in one call — the reorder→PO bridge. Runs the same canonical reorder detection as get_reorder_recommendations, takes its buy-from-vendor bucket, and creates draf
  • list_simulation_scenario_runs read-onlyList the latest 50 scenario runs of the simulation workspace this credential is connected to, newest first: name, status, when it ran, why it failed, and how many expectations passed. Read one with get_simulation_scenari
  • push_external_sales_ordersPush sales orders fulfilled OUTSIDE Spark for an External warehouse. They feed demand velocity without touching on-hand (the snapshot stays the inventory authority). Idempotent: each order is keyed by its stable external
  • set_import_mappingSet (or replace) the declarative column mapping for a staged import session and validate it against EVERY staged row. Author the mapping from the profile get_import_session returns. For salesOrderHistory: fields keyed by
  • get_stock_movements read-onlyAn item's stock movement history at one warehouse, newest first — the Inventory page's movement log: receipts, sales, adjustments, moves, transfers and (with includeReservations) reservations, each with the document, loc
  • explain_demand_change read-onlyWhy an item's demand moved: its last 4 complete weeks of units against the expectation set by the 12 weeks before (direction down / up / flat, or insufficient_data), the weekly series with the same weeks last year, and t
  • get_portfolio_health read-onlyThe 3PL portfolio page: inventory value and overstock across clients, each client's health (score, at-risk SKUs, overstock), the largest overstocked positions, upcoming purchase-order arrivals and predicted reorders, and
  • get_rate_components read-onlyList all rate components on a ComponentPriced price book — keys, labels, basis formulas, rate entries, conditions, and clamps. Read-only. Use to inspect what a book will resolve before activation, or to verify components
  • upsert_item_sourcing_profileCreate or update an item's sourcing profile (its default price book + production spec). Matching: with name, the item's profile of that name is updated (or created); without name, the item's single existing profile is up
  • update_onboarding_intentSet or update this tenant's onboarding INTENT — what they want from Spark: goals, the data they already hold, the sales channels/integrations they plan to use, and free import-scope notes. PATCH semantics: omit a field t
  • list_diagnostic_investigations read-onlyList this account's diagnostic investigations, newest first. Returns an empty list when the feature is unavailable.
  • reset_simulation_workspaceReturn the simulation workspace this credential is connected to to its baseline, the client's data before anything was rehearsed (governed two-step, destructive): rehearsed records, the simulated books' postings and the
  • save_b2b_store_settingsSet up or change the B2B store (see get_b2b_store_settings): its address slug, name, logo, colors, whether it is open, whether browsing needs a sign-in, whether prices are shown and the minimum order. Only the fields you
  • create_supplier_returnSend stock back to a supplier against a closed purchase order — the purchase order's Create supplier return. Each line names the purchase order line (or SKU) and how many units go back, at most what was received minus wh
  • get_receipt_variances read-onlyTenant-wide work queue of unresolved delivery-receipt variances — DR lines where the supplier-delivered cost and/or quantity differs from the sourced purchase order's line snapshot and no resolution decision (approve/rej
  • record_invoice_paymentRecord a payment received outside Spark (cash, check, bank transfer, card taken elsewhere) on an issued invoice — the invoice list's Mark as paid. Nothing is charged: this never runs a card or bank payment. The amount de
  • transition_work_orderMove one work order to a new status: approve (waitingApproval to readyForProduction), mark_ready (open or onHold to readyForProduction), hold (open to onHold), or cancel (any work order not started, completed or cancelle
  • acknowledge_actionPreview or commit acknowledgment of one Signal Center event. Acknowledgment stamps the first-read time and resolves its linked source notification; it does not approve, execute, or dismiss anything. First call without re
  • run_standard_report read-onlyRun one standard report by name (list_standard_reports lists them and the arguments each takes) and return its rows. Pass only the arguments that report lists; any other is refused. Examples: report=top_selling_items wit
  • create_quoteDraft a quote for a customer — the Quotes page's New quote. Lines are priced from the price book for that customer unless you give unitPrice; the preview shows the totals after discounts and tax. The quote is numbered an
  • restore_simulation_snapshotPut the simulation workspace this credential is connected to back exactly as a snapshot holds it, simulated clock included (governed two-step, destructive). Everything done since — rehearsed or imported — is replaced, an
  • issue_purchase_orderIssue a purchase order that is in Created status. An order a person has already approved is issued; an unapproved order is routed to Waiting Approval instead, because an assistant never approves a purchase order. The pre
  • get_item_stock_outlook read-onlyStock outlook for a list of items (by SKU, max 100): available (net of committed, the units open sales and transfer orders hold, as the Inventory page and Shopify show it), committed, inbound, and the two days-of-supply
  • delete_uomDelete a unit of measure by id (GUID) or exact abbreviation, as the Settings page does (a soft delete). Refused while any item or unit conversion uses it, and for the default unit (make another unit the default first wit
  • save_edi_trading_partnerCreate or edit one EDI trading partner. Omit id to create (name required; a new partner is active and in test mode unless set); pass id to edit, where every omitted field keeps its value. Codes are unique per tenant. Con

…and 40 more.

Connect it

Claude Code
claude mcp add --transport http spark-inventory https://api.sparkinventory.com/mcp
Claude.ai / ChatGPT
Settings → Connectors → Add custom connector
URL: https://api.sparkinventory.com/mcp
Cursor (~/.cursor/mcp.json)
{
  "mcpServers": {
    "spark-inventory": {
      "url": "https://api.sparkinventory.com/mcp"
    }
  }
}
VS Code (.vscode/mcp.json)
{
  "servers": {
    "spark-inventory": {
      "type": "http",
      "url": "https://api.sparkinventory.com/mcp"
    }
  }
}

Questions

What does Spark Inventory do?

AI inventory management: stock, Demand ESP reorder recommendations and purchase orders. It exposes 80 tools to AI assistants.

What tools does Spark Inventory provide?

80 tools, including draft_rule, list_standard_reports, delete_manufacturing_order, list_demand_evidence, list_saved_reports, list_edi_item_mappings, start_integration_connect, import_customer_returns, commit_labor_plan, move_stock, get_labor_work_day, create_supplier_price_book….

How do I connect Spark Inventory to Claude, ChatGPT or Cursor?

Add the remote MCP endpoint https://api.sparkinventory.com/mcp as a custom connector or MCP server. Copy-paste setup is on this page.

Does Spark Inventory require an account or API key?

No sign-in was needed to list its tools. Individual tools may still need an account or API key.

Is Spark Inventory online?

As of 2026-11-19 our probe found it online. It answered 100% of our last 3 checks.

Who publishes Spark Inventory?

sparkinventory.com. The official MCP Registry verified that the publisher controls this domain before accepting the listing.

Similar in E-commerce & Shopping