Meilisearch
AI integration·MCP server

Manage Meilisearch from your LLM workflow

Index documents, tune ranking rules, and search Meilisearch through natural conversation in the AI client you already use.

View on GitHubRead the docs
Update the ranking rules for the movies index to prioritize word matches and handle typos.
Calling the Meilisearch MCP server…
Tool · update-settingsok
{
  "index": "movies",
  "rankingRules": [
    "words", "typo", "proximity"
  ]
}
Settings updatedtask #4821

Built for developers using Meilisearch with AI

Manage indexing and relevancy from the same conversation you use to ship features.

Tune relevancy from your LLM

Update ranking rules, searchable attributes, and faceting through natural-language prompts.

Manage indexes and documents

Create indexes, add documents, run searches, and inspect tasks without leaving your AI client.

Fits the workflow you already use

Works in Claude Desktop, Cursor, OpenAI agents, and any MCP-compatible client.

Every tool the server exposes

The server wraps the Meilisearch API as a set of named tools your LLM can call.

Index management

  • ›create-index
  • ›list-indexes
  • ›delete-index
  • ›get-index-metrics

Documents

  • ›get-documents
  • ›add-documents

Search

  • ›search across single or multiple indexes
  • ›filtering and sorting

Settings & relevancy

  • ›get-settings
  • ›update-settings

API keys

  • ›get-keys
  • ›create-key
  • ›delete-key

Tasks & system

  • ›get-task
  • ›get-tasks
  • ›cancel-tasks
  • ›delete-tasks
  • ›health-check
  • ›get-version
  • ›get-stats

Use the MCP client you already have

One server, exposed to every client that speaks the Model Context Protocol.

Claude Desktop

The setup documented in the Meilisearch docs. Add the server to claude_desktop_config.json.

Cursor

Add the server to Cursor’s MCP settings to manage indexes and relevancy from your editor.

OpenAI agents

Listed in the README as a supported client. Wire the server into agent workflows.

Any MCP-compatible client

The server speaks the open Model Context Protocol, so any MCP host can connect to it.

Quick start

Three steps to start chatting with your Meilisearch instance.

  1. 1

    Install the server

    Install
    $ uvx -n meilisearch-mcp

    pip install meilisearch-mcp and a Docker image are also available. See the README.

  2. 2

    Add it to your MCP client

    Example for Claude Desktop. Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

  3. 3

    Tell the assistant to connect

    Please connect to my Meilisearch instance at MEILISEARCH_URL using the API key API_KEY

    From there, ask in plain language: create indexes, update ranking rules, inspect tasks.

Ready to try it?

Spin up a Meilisearch project on Cloud, point the MCP server at it, and start managing search from your AI client.

Try Meilisearch CloudView on GitHub