MCP

What is MCP?

MCP is an experimental Model Context Protocol server that allows AI assistants to interact with Algolia APIs. It is intended for developers who want to query, add, update, or configure Algolia indices through natural language or an MCP-enabled client.

How to use MCP?

Install Go, clone the repository, build the server binary, then add a server definition to Claude Desktop's settings. Configure the environment variables ALGOLIA_APP_ID, ALGOLIA_INDEX_NAME, ALGOLIA_API_KEY, and optionally ALGOLIA_WRITE_API_KEY (the admin key) for write operations. Restart Claude Desktop to access the "algolia" tool.

Key features of MCP

• Search Algolia indices with complex filters.
• Add JSON objects to an index.
• Update searchable attributes and settings.
• Configure ranking and other index options.

Use cases of MCP

• An AI assistant searches for products by brand and price.
• Adding a blog post as a JSON object to an index.
• Updating the searchable attributes for a recipes index.
• Configuring an index to rank award winners higher.

FAQ from MCP

What can I do with this MCP?

You can search, add, update, and configure Algolia indices using natural language prompts, such as “Search all products where brand = ‘Nike’ and price < 100” or “Update the searchable attributes for the recipes index.”

Is MCP officially supported by Algolia?

No. This is an experimental project provided “as is” and “as available,” with no guarantees, warranties, or SLA coverage. It may be updated, broken, or removed at any time.

What are the runtime requirements?

You need Go to build the server binary. The binary runs inside Claude Desktop or via a local MCP host (e.g., mcphost) with Ollama.

Does Claude Desktop support resource templates?

Currently, Claude Desktop does not support resource templates or root, despite online examples showing them. This is a known limitation.

Where can I learn more about this MCP server?

A blog post with an embedded demo video is available at the Algolia engineering blog.