Inventory that restocks itself — until it shouldn’t.
Items, warehouses, stock levels, suppliers, purchase orders, transfers, and cycle counts in one place — with AI agents that reorder low stock, flag what’s gone dead, and catch count variances before they become a write-off.
$ humlens agent run auto-reorder Scanning items below reorder point... ✓ SKU BLW-014 on-hand 8 reorder pt 25 → preferred supplier: Northwind Textiles (lead time 6d) → drafted PO-1042 for 150 units ($1,380.00) → within policy limit ($5,000/day) — awaiting approval ✓ SKU CRD-221 on-hand 0 flagged: dead stock (94d idle) Logged 2 actions to the agent audit trail.
Everything inventory needs, nothing it doesn't.
One workspace for the whole stock lifecycle — items, warehouses, suppliers, purchase orders, transfers, and counts — built around the team that already runs it.
SKUs, categories, unit cost, and reorder points in one place — the source of truth everything else is built on.
On-hand and reserved quantities per item, per warehouse, updated in real time — no spreadsheet reconciliation.
Track lead times and link each item to a preferred supplier, so replenishment never stalls on "who do we order from?"
Issue a PO, then record what actually arrives — partial receipts update stock automatically, line by line.
Move stock between warehouses with a two-step handoff, or correct for damage, loss, or found inventory with a reason code.
Plan a blind count against a snapshot, then reconcile variances automatically once counting is done.
Every receipt, issue, transfer, and adjustment is one append-only record — stock levels can never drift from it.
Set autonomy limits per team — max auto-reorder spend, eligible categories, dead-stock and variance thresholds — before any agent acts.
From catalog to count.
Four steps, one ledger — with an agent watching every movement.
- 1 Track
Catalog every item and see on-hand stock across every warehouse, updated the moment it moves.
- 2 Replenish
When an item hits its reorder point, the agent drafts a PO from the preferred supplier automatically.
- 3 Move
Receive against a PO, transfer between warehouses, or adjust for damage and loss — all in one ledger.
- 4 Reconcile
Run a cycle count and let the agent flag variances beyond tolerance before they become a surprise.
Meet your AI inventory team
Every agent works within a policy you set, and logs a reasoned entry for each decision it makes — or declines to make. Nothing happens silently.
Auto-reorder agent
When an item with a preferred supplier hits its reorder point, drafts a real purchase order automatically — within the limits you set.
Dead-stock agent
Flags items that have sat with no outbound movement past your threshold, before they turn into a write-off.
Cycle-count anomaly agent
Runs the moment a count completes and flags any line whose variance exceeds your tolerance for a second look.
Natural-language stock entry
"Received 50 units of blue widgets at the main warehouse" becomes a structured, ready-to-submit draft.
Also exposed as an MCP server — Claude or any MCP-compatible agent can check stock levels, record a movement, or create a purchase order directly, under the same role-based permissions as a human teammate.
Humlens Inventory is also available self-hosted, on your own infrastructure, with your own AI provider keys — the same BYOC model as the rest of the Humlens platform.
Everything about Humlens Inventory.
- What triggers an AI reorder?
- When an item’s on-hand quantity drops to or below its configured reorder point and it has a preferred supplier assigned, the auto-reorder agent drafts a purchase order for the configured reorder quantity — but only if it falls within your agent policy’s spend and category limits. Every draft is logged with its reasoning and either auto-approved within policy or routed for human approval.
- How do agent policy limits work?
- Each team sets its own AgentPolicy: whether auto-reorder is enabled, a maximum auto-reorder amount, which categories agents are allowed to act on, how many idle days count as dead stock, and the variance tolerance for cycle-count anomalies. Agents can only propose or execute actions inside those limits — anything outside is rejected by policy or surfaced for a human to decide.
- Does it support multiple warehouses?
- Yes. Stock levels, transactions, transfers, and cycle counts are all tracked per warehouse. You can see on-hand and reserved quantities for any item across every warehouse in real time, and move stock between them with a two-step transfer that deducts on send and adds back on confirmed arrival.
- What happens to actions the agent takes?
- Every proposed or executed agent action — an auto-reorder suggestion, a dead-stock alert, a cycle-count anomaly, a natural-language stock entry — is written to an audit log with its input, output, reasoning, and confidence, and its status (proposed, executed, rejected by policy, or overridden by a human). Nothing an agent does happens silently.
- Can I use Humlens Inventory with my own AI agents or tools?
- Yes — inventory data and actions are also exposed as an MCP server, so Claude or any MCP-compatible agent can check stock levels, record a movement, or create a purchase order directly, under the same role-based permissions as a human teammate.
- Is Humlens Inventory available self-hosted?
- Yes. Humlens Inventory is also available self-hosted, on your own infrastructure, with your own AI provider keys — the same self-hosted model as the rest of the Humlens platform. See the self-hosted page for details.
Ready to let inventory manage itself?
Set up your warehouses, add your catalog, and configure exactly how much the agents are allowed to do.