Plug ChatGPT, Claude or Meta Muse into ShelfSpace. Ask it anything about your money and it answers from your real books, then does the work. Run a dispensary? Invoices checked against Metrc, credits recovered, vendors paid. Run a brand? Past-due invoices chased, promises to pay tracked, every invoice proven against Metrc. You never have to open ShelfSpace.
Same feature. Pick the explanation that fits how you think.
MCP server over Streamable HTTP at /api/mcp. Add it as a custom connector.OAuth 2.1 + PKCE (S256), dynamic client registration, client ID metadata documents, and customer-registered confidential clients.POST /api/agent/v1/tools/{tool}, described by OpenAPI 3.1.Dispensaries run payables, Metrc checks and credits. Brands run receivables, collections and proof of delivery. One conversation either way.
See what's due, what's ready and what's blocked. Line up the week's payments and send them within the limits your company sets.
Every invoice, line by line, against the Metrc manifest that moved the product. Your AI explains each mismatch in plain words.
Returns, destroyed product and price drops, rolled into each vendor's monthly credit memo. Ask where every dollar stands.
Ask for a corrected invoice, a missing copy or a credit, with the numbers already in the note. Every message is labeled as written by your AI.
Hand your AI a photo or PDF of an invoice, a COA or a manifest. It lands in the right store for review, tagged as added by your AI.
Who owes you, how much and how late, by store and by sales rep. Your AI ranks it and tells you who to chase first.
From your billing address, with a pay link in every one. Every message is labeled as written by your AI.
When a buyer commits to a date, your AI knows it, and tells you the morning it slips.
Each invoice matched to the buyer's accepted Metrc manifest, so "we never got it" ends the conversation, not starts it.
Payments received, partial payments and payment plans, plus what's due next week across every buyer.
Days to pay by store, by chain and by rep, so credit terms follow the data.
Weekly settlements, splits, returns and what each brand is owed, explained line by line.
Know what synced, what's waiting and why. Catch a stuck bill, invoice or payment before your bookkeeper does.
Deliveries, invoices, line items, purchase orders, payments, credits and sales, filtered and sorted any way you like, then exported to CSV.
No new app to learn. Your AI does the clicking.
In ChatGPT or Claude, add a connector and paste the ShelfSpace address. In Meta Muse, use the app you register in Settings.
Your normal ShelfSpace login and code, one time. Choose which stores your AI may use, then click Allow.
Questions, reports, uploads, vendor notes and payments, all from the chat window you already live in.
MCP for AI assistants, REST for everything else. The same tools, permissions, limits and audit trail behind both.
Your AI gets exactly your access and nothing more, and every move it makes is on the record.
Your login and verification code, once. No shared passwords, no database keys handed to an AI.
Choose stores at connect time. Everything else stays invisible to it, every single call.
A read-only person gets a read-only AI. Permissions can only narrow, never widen.
Have a person confirm each payment, or let your AI pay within per-payment and per-day limits you set.
Each change is logged as done by your AI, with which app, when and what. Your team sees the badge.
Disconnect any AI from Settings and it loses access immediately. Bank details always need a person to confirm.
Standards all the way down. If your client speaks MCP or OAuth, it already speaks ShelfSpace.
https://ourshelf.space/api/mcp/.well-known/oauth-authorization-server (RFC 8414), /.well-known/oauth-protected-resource (RFC 9728)shelfspace.read · shelfspace.actPOST /api/agent/v1/tools/{tool} · GET /api/agent/v1/openapi.json# Settings → Connectors → Add custom connector Name ShelfSpace Server URL https://ourshelf.space/api/mcp Auth OAuth # you'll sign in and pick stores # then just ask: "Which of my stores can you see?"
{
"mcpServers": {
"shelfspace": {
"type": "http",
"url": "https://ourshelf.space/api/mcp"
}
}
}
// OAuth + PKCE runs on first connect
curl -X POST \ https://ourshelf.space/api/agent/v1/tools/list_my_stores \ -H "Authorization: Bearer $ACCESS_TOKEN" \ -H "Content-Type: application/json" \ -d '{}' # Full tool list + schemas: # GET /api/agent/v1/openapi.json
Already a customer? Connect your AI today. New to ShelfSpace? We'll set up your account and you'll be chatting with your books in days.
Yes. Brands on ShelfCollect connect the same way. Your AI can see who owes you and how late, send invoices, statements and reminders, track promises to pay, and show which invoices match the buyer's accepted Metrc manifest.
Yes. Ask who's past due and tell it to send reminders. They go from your billing address with a pay link, and your AI tracks every promise to pay so you know the moment one slips.
Yes. ShelfSpace is a remote MCP server for cannabis back-office finance. Connect ChatGPT, Claude, Meta Muse or any MCP-compatible AI and it can work your accounts payable, receivables, Metrc invoice matching, QuickBooks, credit recovery, consignment settlements and vendor payments, scoped to your own stores.
Yes. Add ShelfSpace as a connector in ChatGPT or Claude, sign in once, and pick your stores. Your AI can then find invoices, check them against Metrc, see what's owed, line up vendor payments, upload new invoices and answer questions about any vendor.
Yes. Register an app in ShelfSpace Settings under AI agents, choose the Meta Muse preset, and paste the Client ID and Client Secret into a Muse custom connector.
Only once. You sign in with your normal login and verification code to connect your AI. After that you work entirely from your AI's chat window.
Yes, once your company turns on AI payments. You choose whether a person confirms each payment or whether your AI pays within per-payment and per-day limits you set. Bank-detail changes always need a person on your team to confirm.
Yes. Your AI can pull every invoice with its Metrc match, line by line, and explain why a line is short, over, missing from the manifest, or billed twice.
Your AI can see what has synced to QuickBooks, what's waiting and why, so you can catch a stuck bill or payment without opening either system.
Your AI signs in as you and only sees the stores you choose. A read-only person gets a read-only AI. It never holds a database password, every action is logged as done by your AI, and you can disconnect it in one click.
MCP (Model Context Protocol) is the open standard AI assistants use to connect to outside tools and data. ShelfSpace speaks it, so your AI can use ShelfSpace the way a person on your team would.
Yes. The same tools are available as a REST API at /api/agent/v1/tools/{tool}, described by an OpenAPI 3.1 file, using the same OAuth token, permissions and audit trail as the MCP server.