Meilisearch MCP Server – Search Index Control

The official Meilisearch MCP Server lets MCP-compatible AI clients interact with a running Meilisearch instance through natural language workflows. It is a stdio-based Python server for managing indexes, documents, search settings, API keys, tasks, health checks, and search operations.

#search#indexes#documents

Overview

Meilisearch MCP Server is the official Model Context Protocol server for connecting AI assistants to Meilisearch. It gives an MCP-compatible client a controlled way to inspect and manage search infrastructure without manually writing every Meilisearch API request. The server is implemented in Python, is distributed as the meilisearch-mcp package, and currently uses the stdio transport.

What the MCP server enables

The server exposes Meilisearch operations as tools. An AI agent can view or update connection settings, create and delete indexes, list indexes, inspect index metrics, add or update documents, retrieve documents with pagination, and run searches. Search support includes filters, sorting, facets, multi-index search, semantic search with vectors, and hybrid keyword plus semantic search where supported by the connected Meilisearch instance.

It also exposes operational tools for getting and updating index settings, managing API keys, checking health, retrieving version and system information, getting database statistics, and monitoring tasks. Task tools include listing tasks, inspecting a task, canceling pending or enqueued tasks, and deleting completed tasks.

When to use it

Use this MCP server when a developer, search engineer, or support workflow needs conversational control over Meilisearch. Practical examples include creating a product index, loading sample or production documents, adjusting ranking rules, configuring searchable and displayed attributes, testing filtered searches, checking whether indexing tasks have completed, and validating the health of a Meilisearch Cloud or self-hosted deployment.

Connection and authentication

The documented Claude Desktop setup starts the MCP server with uvx -n meilisearch-mcp. Meilisearch documents Python 3.9 or newer, a Meilisearch Cloud or self-hosted instance, and the Meilisearch host plus API key from the dashboard. Default connection settings can be supplied with MEILI_HTTP_ADDR and MEILI_MASTER_KEY, or updated during a chat through the server's connection-setting tool. The Docker image can also be run with the same environment variables.

Key considerations

The server currently uses stdio; the project notes that native Meilisearch MCP support is planned for later. Because the server can create and delete indexes, mutate documents, update settings, and manage API keys, credentials should be scoped carefully and treated as sensitive. Meilisearch specifically warns that updating hosts and API keys directly in chat is convenient for development and multi-instance testing but does not follow best MCP security practices for production unless additional safeguards are in place.

Frequently Asked Questions

When should an AI agent use the Meilisearch MCP Server?

Use it when an agent needs to inspect, configure, or operate a Meilisearch instance through an MCP-compatible client. Typical workflows include creating indexes, loading documents, tuning settings, testing searches, monitoring indexing tasks, and checking instance health.

What does this MCP server add to an AI agent's capabilities?

It gives the agent live access to Meilisearch tools instead of relying on static knowledge. The agent can call tools for connection settings, indexes, documents, searches, settings, API keys, tasks, health, version, statistics, and system information.

What can the AI agent access or manage through this server?

The documented tools cover connection management, index creation/listing/deletion, index metrics, document retrieval and updates, flexible search, index settings, API key operations, task management, health checks, version details, database statistics, and system information.

How is authentication configured for this MCP server?

The server connects to a Meilisearch instance using a host URL and API key. These can be provided as `MEILI_HTTP_ADDR` and `MEILI_MASTER_KEY`, or the connection can be updated through the MCP server's connection-setting tool. Secrets should be kept out of chat history in production workflows.

Which transport should be used for this MCP server?

Use the stdio transport. The official README describes the server as stdio-based and shows `uvx -n meilisearch-mcp` for MCP clients, while Docker can be used to run the same stdio server in containerized environments.