Algolia MCP Server

Experimental MCP server for Algolia: run searches, fetch recommendations, manage indices, and pull analytics from your Algolia applications.

Content & CMS by Algolia API Key active

GitHub Docs

Overview

The Algolia MCP Server exposes Algolia's search platform to LLM clients through the Model Context Protocol. It lets an AI agent search and write to indices, retrieve recommendations and query suggestions, run A/B tests, and pull analytics, monitoring, and usage metrics from your Algolia applications. Algolia maintains two open-source reference implementations: a Go server at algolia/mcp and a Node.js server at algolia/mcp-node that integrates with the Algolia Dashboard via browser-based authentication.

The Go server exposes a focused set of tool groups: abtesting, analytics, collections, monitoring, querysuggestions, recommend, search (read + write), search_read, search_write, and usage. Each group can be enabled or disabled via the MCP_ENABLED_TOOLS environment variable, which is useful for scoping a client to read-only operations or restricting it to a single index. Authentication uses standard Algolia application credentials (App ID, search API key, and optionally an admin write key).

Algolia also offers hosted MCP options described on their developer site: a Public MCP for application-scoped search and recommendations and a Productivity MCP for user-scoped, read-only analytics across all accessible applications. Both servers are labeled experimental and are provided without SLA, but they are maintained under the official algolia GitHub organization.

Tools

Tool	Description
`search`	Combined read and write search operations against an Algolia index (queries, settings, object CRUD).
`search_read`	Read-only search operations: list indices, get settings, run queries, fetch objects.
`search_write`	Write operations: clear, copy, delete, move indices, set settings, insert or delete objects.
`recommend`	Fetch Algolia Recommend results such as related products and frequently bought together items.
`querysuggestions`	Manage and retrieve Query Suggestions configurations.
`analytics`	Pull analytics such as top searches, no-result rates, and trending queries.
`abtesting`	Create, list, and inspect A/B tests on your indices.
`monitoring`	Check application latency, incidents, and infrastructure status.
`collections`	Manage Algolia merchandising collections.
`usage`	Retrieve account usage metrics such as operations and record counts.

Setup Guide

Prerequisites

Go installed locally (for building the Go server) or Node.js 22+ for algolia/mcp-node
An Algolia account with an Application ID and API keys

Build the server

git clone git@github.com:algolia/mcp.git
cd mcp/cmd/mcp
go build

Note the resulting binary path.

Claude Desktop configuration

Add the following to your Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "algolia": {
      "command": "/path/to/the/repo/cmd/mcp/mcp",
      "env": {
        "ALGOLIA_APP_ID": "<APP_ID>",
        "ALGOLIA_INDEX_NAME": "<INDEX_NAME>",
        "ALGOLIA_API_KEY": "<API_KEY>",
        "ALGOLIA_WRITE_API_KEY": "<ADMIN_API_KEY>",
        "MCP_ENABLED_TOOLS": "",
        "MCP_SERVER_TYPE": "stdio",
        "MCP_SSE_PORT": "8080"
      }
    }
  }
}

Environment variables

ALGOLIA_APP_ID: Application ID from the Algolia dashboard
ALGOLIA_INDEX_NAME: Default index name to scope operations
ALGOLIA_API_KEY: Search API key for read operations
ALGOLIA_WRITE_API_KEY: Admin API key, required only for write tools
MCP_ENABLED_TOOLS: Comma-separated list of tool groups to enable (defaults to all). Valid values include abtesting, analytics, collections, monitoring, querysuggestions, recommend, search, search_read, search_write, usage
MCP_SERVER_TYPE: stdio (default) or sse
MCP_SSE_PORT: Port for SSE mode (default 8080)

Restart Claude Desktop after editing the config. The Node.js variant (algolia/mcp-node) instead uses a browser-based authenticate command to sign in with your Algolia Dashboard credentials and offers a list-tools CLI command to inspect available tools.

Use Cases

Run product searches against an Algolia index and have the agent summarize results or filter facets in natural language
Pull "top searches with no results" and trending queries to feed merchandising or SEO workflows
Insert, update, or delete records in a catalog index from an agent driven import or content sync
Inspect A/B test results and recommend a winner based on conversion or click-through rates
Monitor application latency and incidents and surface alerts to a Slack or email workflow

Example Prompts

"Search the products index for waterproof hiking boots under $200 and return the top 5 hits."
"Show the no-results queries in the last 7 days and group them by frequency."
"Get Recommend related-products results for objectID SKU-12345 in the products index."
"List active A/B tests and tell me which variant is winning by conversion rate."
"Check current latency and any open incidents for application ABC123."

Pros

Maintained under the official algolia GitHub organization with both Go and Node.js implementations
Tool groups can be scoped via MCP_ENABLED_TOOLS to restrict the agent to read-only or analytics-only access
Covers the full surface of Algolia: search, Recommend, A/B testing, query suggestions, analytics, monitoring, and usage
Supports both stdio and SSE transports

Limitations

Labeled experimental and provided "as is" with no SLA or formal support
The Go server requires building from source; no prebuilt binary or npm package is provided for it
Tool descriptions in the README are high level; precise parameter schemas require reading the source or running list-tools

Alternatives

Meilisearch MCP for a similar search-engine integration
Elastic MCP server for Elasticsearch indices
Algolia's hosted Public MCP and Productivity MCP described at algolia.com/developers/lp-mcp if you prefer a managed endpoint