Oyster MCP server

Put Oyster inside Claude, OpenAI and Gemini.

Oyster speaks the Model Context Protocol, so any assistant that supports MCP or tool use can run a skin scan, pull recommendations from a catalogue and read skin intelligence, in plain language.

Book a demo ↗Read the docs
MCP + tool useAny client
29 countriesLive today
Every skin toneAccurate
assistant · oyster
# an assistant, connected to Oyster
> Scan this customer photo and build
  a routine from our catalogue.

→ oyster.run_skin_scan(image)
  read across 3000+ concerns,
  accurate on every skin tone
→ oyster.match_routine(scan_id)
  4 products matched · confidence 0.92
Plain language
Real skin intelligence
What you can do

Skin analysis tools your assistant can call.

Oyster publishes its capabilities as MCP tools. An assistant calls them the way it calls any other tool, in one conversation. Every scan runs on Oyster's proprietary imaging method, read across 3000+ concerns, with a human aesthetician in the loop.

run_skin_scan

Analyse a face from an image

Send a photo and get a full skin analysis back, read across 3000+ concerns and accurate on every skin tone.

match_routine

Build a routine from a catalogue

Turn a scan into a matched routine drawn from a connected catalogue, ranked and ready to recommend.

product_matches

Look up matches and confidence

Ask which products fit a concern or a scan, each returned with a confidence score you can act on.

concern_trends

Query trends and analytics

Read macro skin concern trends by region and city. No individual identity is ever exposed in the numbers.

catalogue_items

Manage catalogue items

Create, update and read catalogue entries, so recommendations only ever point at products you actually stock.

order_lookup

Check an order or attribution

Look up an order and the scan it was attributed to, and read the intelligence behind a completed sale.

Three assistants, one server

Connect once. Run it from Claude, OpenAI or Gemini.

Connect your assistant to Oyster with an API key and permission to use the tools it needs. The examples below show the connection format and use placeholder keys.

Claude

Claude Desktop and API

Add the Oyster server to your MCP servers config in Claude Desktop, or wire it into the Claude API through agent tools.

claude_desktop_config.json
{
  "mcpServers": {
    "oyster": {
      "url": "https://mcp.oysterskin.com",
      "headers": {
        "Authorization": "Bearer OYSTER_API_KEY"
      }
    }
  }
}
OpenAI

Function calling and Responses

Register Oyster as a tool through the Responses API, or reach the same server over an MCP bridge in your agent stack.

responses.py
# OpenAI Responses API · MCP tool
client.responses.create(
    model="gpt-5",
    tools=[{
        "type": "mcp",
        "server_label": "oyster",
        "server_url": "https://mcp.oysterskin.com",
        "headers": {"Authorization": "Bearer OYSTER_API_KEY"},
    }],
    input="Scan this photo and match a routine.",
)
Gemini

Gemini function calling

Declare Oyster as a function in the Google GenAI SDK and let Gemini call it during tool use.

gemini.py
# Gemini · function declaration
oyster = {
    "name": "oyster_skin_scan",
    "description": "Run an Oyster scan, return a routine.",
    "parameters": {"type": "object",
        "properties": {"image_url": {"type": "string"}}},
}

client.models.generate_content(
    model="gemini-2.5-pro",
    contents="Scan this face and recommend a routine.",
    config={"tools": [{"function_declarations": [oyster]}]},
)
How it connects

From key to first answer in four steps.

Connect an assistant that supports MCP or tool use. Oyster runs the skin analysis, so you do not need to host the model.

1

Get an API key

Create a key in the Oyster dashboard and choose the scopes it carries.

2

Connect your assistant

Aim your client at mcp.oysterskin.com and pass the key as a bearer token.

3

Choose its permissions

Allow only the tools you need, from a read only scan to full catalogue writes.

4

Start asking

Ask in plain language. The assistant calls the tools and reads the results back.

Security

Access is scoped, private and certified.

Scoped keysEach key allows access only to the tools you selected.
Short lived sessionsTool sessions expire after a short period rather than keeping access open.
No identity in analyticsTrend data is macro only. No individual is ever exposed.
ISO 27001 certifiedOyster is independently ISO 27001 certified.
FAQ

The Oyster MCP server, answered.

What is the Model Context Protocol?+
MCP is an open standard for connecting AI assistants to outside tools and data. A server publishes its capabilities, and any assistant that speaks the protocol can discover and call them. Oyster runs an MCP server, so an assistant can invoke a skin scan, a routine match or an analytics query as native tools.
Which assistants can connect to Oyster?+
Claude, OpenAI and Gemini, plus any other client that supports MCP or tool use. The same Oyster server answers to all of them. You connect it once and reach the same tools from whichever assistant your team runs on.
Do I need to be technical to use this?+
A developer sets up the connection and API key. After that, your team can ask questions in plain language. Book a demo if you would like help connecting your assistant.
Is my data safe?+
Keys are scoped to the tools you grant, sessions are short lived, and no individual identity is exposed in any analytics the tools return. The platform is ISO 27001 certified, and the API key stays on your side and is never embedded in a client that reaches a browser.
How do I get access?+
Book a demo and we provision a scoped Oyster API key for your account. Read the developer docs for the full tool reference and connection steps, then point your assistant at the server and start asking.
Get started

Give your assistant real skin intelligence.

Book a walkthrough using your own tools, or follow the developer docs to connect Oyster.

Book a demo ↗Read the docs