About GemstoneAI
GemstoneAI is a hosted MCP server for assistants helping people identify and research jewelry from photos. It returns preliminary analysis, photographed market comparables, and can save a piece to the owner's collection.
Key features
- 01Analyze 1–5 jewelry photos
- 02Show comparable listings and sales
- 03Save pieces to a collection
- 04Report identification alternatives and unknowns
Use cases
- 01Research an owned jewelry piece from photos
- 02Compare a piece with photographed market examples
- 03Save an analyzed piece for access in the app
MCP support
Exposes jewelry analysis, analysis-status, comparable-search, and collection-saving tools to MCP clients. It connects as a remote Streamable HTTP server and uses OAuth sign-in.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://api.gemstoneaiapp.com/mcpclaude mcp add --transport http com-gemstoneaiapp-gemstoneai https://api.gemstoneaiapp.com/mcp{"mcpServers":{"com-gemstoneaiapp-gemstoneai":{"type":"http","url":"https://api.gemstoneaiapp.com/mcp"}}}Open in Cursor{"servers":{"com-gemstoneaiapp-gemstoneai":{"type":"http","url":"https://api.gemstoneaiapp.com/mcp"}}}Open in VS Code{"mcpServers":{"com-gemstoneaiapp-gemstoneai":{"type":"http","url":"https://api.gemstoneaiapp.com/mcp"}}}{"mcpServers":{"com-gemstoneaiapp-gemstoneai":{"serverUrl":"https://api.gemstoneaiapp.com/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEGemstoneAI for Agents
GemstoneAI is a hosted remote MCP server that lets an AI assistant help someone with a piece of jewelry they own. The assistant can:
- analyze 1 to 5 photos of the piece and give a preliminary identification, with its alternatives, unknowns and the evidence that would settle them
- show photographed comparable listings and sales, each labelled as an asking price or a realized sale
- save the piece to the person's GemstoneAI collection, so it opens in the GemstoneAI app on their phone
https://api.gemstoneaiapp.com/mcp
There is nothing to install or run. Product page: gemstoneaiapp.com/agents.
Connect
Add the URL as a remote (Streamable HTTP) MCP server in any client that supports one. The client discovers sign-in on its own. The person signs in with the email they use in the GemstoneAI app, so the collection is the same; a new email gets a free account. Each new analysis uses a scan from that account's plan, the same as in the app.
Claude Code, as a plugin (adds the server and a jewelry-research skill):
/plugin marketplace add Iveye-Inc/gemstoneai-agents
/plugin install gemstoneai@gemstoneai
Claude Code, server only:
claude mcp add --transport http gemstoneai https://api.gemstoneaiapp.com/mcp
Then run /mcp in Claude Code to sign in.
Other clients (Claude on the web and desktop, ChatGPT developer mode, Cursor, VS Code, MCP Inspector): add a remote MCP server with the URL above and sign in when prompted.
Client support is listed on the product page. A client is marked tested only after the full journey (photo, analysis, comparables, save, reopen in the app) has passed in it.
Tools
| Tool | What it does | Annotations |
|---|---|---|
analyze_jewelry | Preliminary analysis of 1 to 5 photos of one piece. Takes photos (attached files) or image_urls (public https), and an optional idempotency_key. With no photos it returns an upload link the person opens on their phone. | writes, not destructive, idempotent, open world |
get_jewelry_analysis | Status or completed result of an analysis. Waits up to about 40 seconds while it is processing. | read-only |
find_jewelry_comparables | Photographed market listings and sales with source, link, price, price type, date found and the material differences from this piece. | read-only, open world |
save_collection_item | Saves the analyzed piece to the signed-in person's collection and returns a link that opens it in the app. | writes, not destructive, idempotent |
Results never carry numeric confidence. Uncertainty comes as alternatives, unknowns and next evidence. When GemstoneAI's own reads disagree about what the stone is, price context is withheld with the reason. When no priced comparables are found, the answer is "none found", never a value of zero.
Protocol and authentication
- Streamable HTTP, stateless, JSON responses. Initialize-based clients (protocol 2024-11-05 to 2025-11-25) and 2026-07-28 clients are supported.
- OAuth 2.1 with PKCE (S256). Discovery:
/.well-known/oauth-protected-resource/mcpand/.well-known/oauth-authorization-serveronapi.gemstoneaiapp.com. - Clients register with Client ID Metadata Documents or dynamic client registration. Loopback redirects may use any port.
- Tokens are bound to one person, one client and this server (RFC 8707). Access tokens last an hour; refresh tokens rotate.
- Scopes:
jewelry:analyzeandcollection:write.
Latency and limits
- An analysis typically takes 20 to 45 seconds.
analyze_jewelrywaits up to about 40 seconds, then returnsprocessingwith a retry interval. Comparables usually follow within a minute. - Repeating the same photos, or the same
idempotency_key, returns the same analysis and is not charged twice. - During the pilot there are daily evaluation limits per account and per integration. Developers can request pilot access on the product page.
What it is not
Results are preliminary analyses from photographs. They are not appraisals and do not establish authenticity, treatment, origin or certified value. The server never buys, sells, lists or contacts anyone, and it saves to a collection only when asked.
Privacy
Photos are used to analyze the piece and are kept with it in the person's GemstoneAI account, like a scan in the app. Location data is removed from every photo. See the Privacy Policy and Terms.
Repository layout
.claude-plugin/marketplace.json,plugins/gemstoneai/: the Claude Code plugin (MCP server config and the jewelry-research skill)openai/: the ChatGPT plugin package source;python scripts/build_openai_zip.pybuilds the ZIPserver.json: the entry in the official MCP Registry, published withscripts/publish_registry.py
Support
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.