Your Inventory,
Understood by AI
AssetBlaze's MCP server gives Claude, ChatGPT, and any AI agent structured, secure access to your inventory — so it can act on your behalf, not just answer questions.
AssetBlaze MCP responds
- ✓ 4 SKUs below threshold
- ✓ matched 4 suppliers
- ✓ PO-1042 drafted — Acme Supply $2,340
- ✓ PO-1043 drafted — NovaParts $870
- ✓ PO-1044 drafted — GreenStock $1,105
- ✓ PO-1045 drafted — Brightfield $440
Ready for your review. Approve to send?
What is MCP?
The protocol that makes AI actually useful for operations
Model Context Protocol is an open standard developed by Anthropic. It gives AI assistants a secure, structured channel to read and act on data in external systems — without screen-scraping, brittle prompts, or handing over your credentials.
Think of it as a typed, permissioned API layer that AI models natively understand. Instead of guessing what your inventory system can do, the AI knows — and can take action safely.
purchase_orders.create
sales_orders.list
suppliers.get_catalog
locations.transfer_stock
How it works
Three steps to an
AI-powered inventory
Add one line to your Claude Desktop or MCP-compatible client config. No SDK installation, no webhook setup, no API key juggling. It takes under two minutes.
Grant read-only or read-write access per resource type. Restrict by location, team, or workflow. Your data never leaves AssetBlaze's infrastructure, and no credentials are shared with the AI.
Talk to your inventory the way you'd talk to a colleague. The AI queries stock levels, creates drafts, flags anomalies, and surfaces insights — all within the permissions you've set.
What you can do
Real workflows, not demos
These aren't hypothetical examples. They're the exact queries AssetBlaze's MCP server handles today.
The agent queries reorder alerts, matches each SKU to its preferred supplier from your catalogue, calculates order quantities, and creates draft POs — all in one turn.
PurchasingNo more pulling CSV exports. The agent reads historical movement data directly, identifies statistical anomalies, and writes a plain-English summary with action recommendations.
AnalyticsThe agent queries current stock, accounts for committed inventory on open orders, checks inbound POs, and gives you a clear available-to-promise number — in seconds.
Sales opsThe agent analyses burn rates per location, models 30-day demand, calculates the optimal inter-warehouse transfer quantities, and creates the transfer orders for your approval.
WarehouseThe agent surfaces slow-moving inventory ranked by capital tied up, checks supplier return policies from your catalogue, and drafts a liquidation recommendation with financial impact estimates.
FinanceDefine natural-language rules that run on a schedule. The agent monitors your inventory, applies your conditions, and surfaces drafts for approval — without you touching the system.
AutomationSecurity model
The AI can only do what you allow
Every MCP request is authenticated, scoped, and logged. Your data stays in AssetBlaze's infrastructure. Nothing is ever sent to an AI model without your explicit permission grants.
AI assistants receive scoped tokens, not your account credentials. Tokens can be revoked instantly from your AssetBlaze dashboard without affecting your main account.
Grant read-only or read-write access per resource type — inventory, orders, suppliers, locations. A reporting agent can never accidentally trigger a purchase order.
Every tool call made by every AI agent is logged with timestamp, user context, and outcome. Replay any session, review agent actions, and spot issues before they become problems.
State-changing operations (create, update, delete) return drafts by default. Nothing is committed to your inventory without an explicit human approval step.
Every mutating endpoint is idempotency-keyed. If an AI agent retries due to a network timeout, you'll never end up with duplicate purchase orders or double-counted stock adjustments.
AssetBlaze's MCP server returns structured responses — not raw database exports. Your supplier pricing, customer data, and stock levels are never bulk-transferred to a third-party AI.
Quick start
Live in 2 minutes
Add AssetBlaze to Claude Desktop or any MCP-compatible AI client with a single config block. No code, no webhooks, no SDK.
- 1Get your MCP token
In AssetBlaze, go to Settings → API & Integrations → MCP. Generate a scoped token and choose your permission level.
- 2Add to your AI client config
Paste the config block into your
claude_desktop_config.jsonor equivalent file. - 3Restart and start asking
Relaunch your AI client. AssetBlaze will appear as a tool provider. Start with "What's my current stock level for…" and go from there.
{
"mcpServers": {
"assetblaze": {
"command": "npx",
"args": [
"-y",
"@assetblaze/mcp-server"
],
"env": {
"ASSETBLAZE_TOKEN": "sk-ab-your-token-here",
"ASSETBLAZE_ORG": "your-org-slug"
}
}
}
}# Read operations inventory.query_stock inventory.query_reorder_alerts inventory.movement_history sales_orders.list purchase_orders.list suppliers.get_catalog locations.list_stock # Write operations (draft mode by default) purchase_orders.create_draft sales_orders.create_draft inventory.adjust_stock locations.create_transfer
Common questions
What people ask before they set it up
MCP is an open standard. It works with any client that supports it — Claude Desktop, Claude.ai, and any custom agent built with the MCP SDK. OpenAI compatibility is on our roadmap.
Both are possible, and you control which. Write operations create drafts by default and require explicit human confirmation. You can lock your token to read-only if you prefer.
MCP access is included on all paid AssetBlaze plans. Read-only access is available on the free plan so you can evaluate it without upgrading first.
Only the structured response from a specific tool call is included in the AI's context. We never bulk-export your data. Your AI provider only sees the data relevant to each specific query.
The REST API is designed for developers writing code. MCP is designed for AI agents operating autonomously — with semantic tool descriptions, structured error hints, and idempotency built in from the ground up.
Draft mode means all write operations are queued for your review, not committed automatically. You approve before anything reaches your suppliers. Idempotency keys also prevent duplicate operations from retry loops.
Your inventory deserves an AI that can act on it
Start free. Connect your MCP client in two minutes. No credit card, no implementation project.
Download
