Algolia MCP Server

The Algolia MCP servers connect AI assistants to the Algolia search and recommendation platform. Anyone using Algolia for search in an online shop, a help center, or an application can ask the assistant questions like "Which search queries return no results?" or "Show me products with this facet value" without opening the dashboard. According to the documentation there are three offerings, all managed by Algolia itself: Public MCP, Productivity MCP, and DevEx MCP. They address different audiences and can be set up separately.

What is the Algolia MCP server?

The Algolia MCP server is not a single program but a family of hosted remote servers. Nothing has to be installed locally: the assistant talks to the respective address over HTTP. That is convenient, but it also means availability lies with Algolia. We found no open-source variant from the vendor; a community project that shows up in directories states in its own description that it is explicitly not supported by Algolia, and it is not described here.

Public MCP: search for customer-facing assistants

The Public MCP makes selected indices and Recommend models available to external assistants, such as a product advisor or a support bot. In the dashboard you create up to five server instances per application and select up to five indices for each. According to the documentation the address follows the pattern https://{APP_ID}.algolia.net/mcp/1/{UNIQ_ID}/mcp. Access is strictly read-only; the client needs no sign-in of its own, and the permissions derive from the search API key. The tools include one search tool per index, facet value search, and recommendations. Every call counts as normal Algolia usage.

Productivity MCP: analysis for staff

The Productivity MCP at https://mcp.algolia.com/mcp is meant for internal use. It is enabled once per account in the dashboard, after which each person signs in with their own login and access follows their Algolia permissions. Besides search and recommendations there are analytics tools, for example for the top searches, searches without results, the no-results rate, filters, and user counts. This offering is read-only as well. Because everything the signed-in person may see is visible, prompts and outputs should be treated as internal data.

DevEx MCP: code instead of data

The DevEx MCP at https://mcp.algolia.com/1/snippets/mcp needs no sign-in and does not access your data. Through the tools algolia_snippets_list_operations and algolia_snippets_get_snippet it returns code examples for the Algolia APIs in the API client languages, matched to the client version. This helps coding agents use current methods instead of outdated ones such as initIndex. The snippets are starting code with placeholders and must be adapted; the server does not generate InstantSearch widgets.

Setup

For Claude Code a single command with the HTTP transport is enough for the DevEx MCP. According to the documentation Cursor, VS Code, ChatGPT, and Claude can be connected through their configuration file or as a custom connector. The Productivity MCP additionally supports Codex and the Gemini CLI. When signing in, leave the OAuth client ID and secret empty.

Security and limits

Public and Productivity MCP only read, so an assistant cannot change indices or settings. Still, search results and analytics data enter the language model's context and thereby reach the model provider. Release only indices to Public MCP instances whose content may be public, and use a search API key with narrow permissions. Write and admin keys never belong in a frontend.

Costs

According to Algolia, using the Public MCP counts toward your existing plan; the documentation names no separate MCP fee. It gives no plan details for Productivity and DevEx. Algolia offers a free entry plan, with paid plans billed monthly above it and an enterprise tier on an annual contract. Current terms are on the official pricing page.

Who is the Algolia MCP server for?

It is worthwhile for teams that already run Algolia: shop operators putting an advisory assistant on their catalog, product teams who want to evaluate search analytics conversationally, and developers who need correct client code. Anyone not using Algolia gains nothing from it, because the servers are an access path to the platform and not a standalone search.

For Public and Productivity MCP an Algolia account with an application and index and an MCP client that supports remote servers; the DevEx MCP needs only the client.

Algolia offers three hosted MCP servers. This guide helps you choose and walks through the setup.

Choose the right offering

The Public MCP is meant for customer-facing assistants, the Productivity MCP for staff who evaluate data and analytics, and the DevEx MCP for coding agents that need API code. According to the documentation all are managed by Algolia.

Step 1: Set up the DevEx MCP in Claude Code

Run claude mcp add --transport http algolia-devex https://mcp.algolia.com/1/snippets/mcp. No sign-in is needed. State the API client version in your prompt, otherwise outdated methods such as initIndex may show up.

claude mcp add --transport http algolia-devex https://mcp.algolia.com/1/snippets/mcp

Step 2: Enable the Productivity MCP

Enable the offering once in the dashboard under Generative AI, MCP Servers, Productivity. Then enter https://mcp.algolia.com/mcp in the client and sign in with your Algolia login. Leave the OAuth client ID and secret empty.

Step 3: Create the Public MCP

Open the MCP Servers page of your application in the dashboard, create a server, and select up to five indices plus any Recommend models you want. Copy the server URL and enter it in the client, in VS Code with the type http.

Step 4: Check permissions and privacy

Release only indices that may be public and use a search API key with narrow permissions. Treat output from the Productivity MCP as internal data, since it reaches the model provider.

Step 5: First requests

With the Productivity MCP ask, for example, "Which searches returned no results last week?". With the DevEx MCP ask for a snippet for a specific operation in your language and check the placeholders.

DevEx MCP without sign-in; Productivity MCP with a personal Algolia login (leave the OAuth client ID and secret empty); Public MCP without client sign-in, permissions via the search API key.

Read-only access to released indices, recommendations, and analytics; Public MCP only to selected indices, Productivity MCP within the permissions of the signed-in person.

The servers run hosted by Algolia. Retrieved search results and analytics data enter the AI client context and therefore reach the model provider.

Releasing non-public indices through the Public MCP, an overly broad search API key, internal analytics data in the model provider's context, and admin keys in a frontend.