Get started with Algolia Public MCP

Use Algolia Public MCP to expose selected indices and Recommend models to external or customer-facing AI assistants. For user-scoped access to the Algolia applications and indices allowed by each user’s permissions, use Algolia Productivity MCP. To compare both, see the MCP overview. Algolia Public MCP provides read-only access. It can search and recommend, but it can’t create, update, or delete indices, records, or settings.

Before you begin

To connect Algolia Public MCP, make sure you have:

An Algolia with at least one and some . if needed.
An MCP client such as OpenAI Playground, ChatGPT, or Claude.

Create or locate an MCP server

Go to the in the Algolia dashboard and select the appropriate application.
If an existing MCP server is available, click the link icon to copy the URL.
To create a :
1. Select up to five indices and any Recommend models you want to expose.
2. Add descriptions to help the LLM choose the right index or tool.
3. Save and copy the URL of the new MCP instance.

You can create up to 5 MCP server instances per application. The MCP server uses your Search permissions to control data access. Your MCP client doesn’t require separate authentication.

Connect to an MCP client

Cursor
VS Code
ChatGPT
OpenAI Playground
Claude

Open Cursor settings, then click Tools and MCPs.
Click Add MCP Server and fill in the following:
- MCP Server URL: https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp
- Name: Algolia
- Description: add a description that helps the LLM understand the tool’s purpose.
- Authentication: none.
Click Save.
Enable the Algolia MCP connection from the prompt bar.

Alternatively, you directly add the following to your ~/.cursor/mcp.json or %USERPROFILE%\.cursor\mcp.json file:

{
  "mcpServers": {
    "Algolia": { 
      "url": "https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp"
    } 
  }
}

Add the following to your ~/.vscode/mcp.json or %USERPROFILE%\.vscode\mcp.json file:

{
  "servers": {
    "Algolia": { 
      "type": "http", 
      "url": "https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp"
    } 
  }
}

Open Connectors settings, then click Create.
Fill in the following:
- Server URL: https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp
- Label: Algolia
- Description: add a description that helps the LLM understand the tool’s purpose.
- Authentication: none.
Check I trust this application.
Click Create.
Enable the Algolia MCP connection from the prompt bar.

Open OpenAI Platform → Playground.
Go to Tools → + Add.
Choose MCP server.
Click + Server and enter:
- Server URL: https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp
- Label: Algolia
- Description: add a description that helps the LLM understand the tool’s purpose.
- Authentication: none.
Click Connect.
Use a system prompt that preserves result fidelity, such as:

You are an AI assistant that helps users with searching for results in an Algolia index.
Display the Algolia results exactly as returned. Do not filter, reorder, or summarize the hits.

Open Connectors settings, click +, then click Add custom connector.
Fill in the following:
- Remote MCP server URL: https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp
- Name: Algolia
- Leave Advanced settings empty
Click Add.
Enable the Algolia MCP connection from the prompt bar.

Restrict access with a different API key per user

To limit each user to their own records, send a secured API key in the x-algolia-api-key HTTP header. The MCP server uses this key instead of the stored API key.

claude mcp add --transport http algolia https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp \
  --header "x-algolia-api-key: ALGOLIA_API_KEY"

{
  "mcpServers": {
    "Algolia": {
      "url": "https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp",
      "headers": {
        "x-algolia-api-key": "ALGOLIA_API_KEY"
      }
    }
  }
}

{
  "servers": {
    "Algolia": {
      "type": "http",
      "url": "https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp",
      "headers": {
        "x-algolia-api-key": "ALGOLIA_API_KEY"
      }
    }
  }
}

from openai import OpenAI

client = OpenAI()

response = client.responses.create(
    model="gpt-5",
    tools=[
        {
            "type": "mcp",
            "server_label": "algolia",
            "server_url": "https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp",
            "headers": {"x-algolia-api-key": "ALGOLIA_API_KEY"},
            "require_approval": "never",
        }
    ],
    input="Find running shoes under $100",
)

print(response.output_text)

Requests without the header use the stored API key. To make the header required, delete the stored API key.

Generate your users’ secured API keys from a different API key. Deleting a key also invalidates the secured API keys generated from it.

Tools exposed by Algolia Public MCP

Algolia Public MCP exposes a tool per index and a small set of shared tools:

Tool	API
Search
`algolia_search_{index_name}`	Search an index
`algolia_search_for_facet_values`	Search for facet values
Recommend
`algolia_recommendations`	Retrieve recommendations

ChatGPT requires two extra tools (see OpenAI MCP docs):

Tool	API
`search`	Search an index
`fetch`	Get an object

Guidelines

Usage and billing. API requests made through Algolia Public MCP count toward your Algolia usage. Each tool call typically maps to one API request.
Data access and permissions. The MCP grants the connected LLM access to your Algolia records according to the permissions defined in your Search .
Prompt design. Use a clear system prompt that tells the assistant to return Algolia results exactly as provided.

Search and Discovery platform

Optimization and Personalization

AI-powered experiences

Production and scale

Get started with Algolia Public MCP

Before you begin

Create or locate an MCP server

Connect to an MCP client

Restrict access with a different API key per user

Tools exposed by Algolia Public MCP

Guidelines

​Before you begin

​Create or locate an MCP server

​Connect to an MCP client

​Restrict access with a different API key per user

​Tools exposed by Algolia Public MCP

​Guidelines

Before you begin

Create or locate an MCP server

Connect to an MCP client

Restrict access with a different API key per user

Tools exposed by Algolia Public MCP

Guidelines