Meilisearch MCP Server
OfficialManage and search Meilisearch from any MCP client: indices, documents, settings, keys, and multi-index search.
Add to your client
Copy the config for your MCP client and paste it into its config file.
pip install meilisearch-mcpPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"meilisearch-mcp-server": {
"command": "uvx",
"args": [
"-n",
"meilisearch-mcp"
],
"env": {
"MEILI_HTTP_ADDR": "http://localhost:7700",
"MEILI_MASTER_KEY": "<your-master-key>"
}
}
}
}Requires `uv` (the Python package runner). Install it from https://docs.astral.sh/uv/ if `uvx` is not found.
Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- A running Meilisearch instance, e.g. docker run -d -p 7700:7700 getmeili/meilisearch:v1.28
- Python >= 3.9 with pip, or uv/uvx (recommended)
- Optional MEILI_MASTER_KEY if your instance requires authentication
About Meilisearch MCP Server
This server turns Meilisearch administration into conversation. The tool list mirrors the Meilisearch API surface: index management (create-index, list-indexes, delete-index, get-index-metrics), document operations (add-documents, get-documents with pagination), a flexible search tool that spans single or multiple indices with filters and sorting, settings management, API key management, task management (get/cancel/delete tasks), and a set of health/stats/version monitoring tools.
Search capabilities include what makes Meilisearch distinctive: typo-tolerant keyword search, faceting, semantic vector search, and hybrid keyword+semantic queries. An agent can build an index, load documents, tune searchable attributes and ranking rules, then iterate on relevance — the full loop that normally means bouncing between docs and curl.
Connection management is dynamic: get-connection-settings and update-connection-settings let the agent inspect and switch between Meilisearch instances mid-session. That is deliberately convenient for development (the README flags it as such) and is the main thing to lock down in shared environments, since URL and key changes are one tool call away.
It is Python ≥3.9, tested against a real Meilisearch in CI, and released to PyPI automatically on version bumps. If you prefer TypeScript, the README points to a community meilisearch-ts-mcp; and since Meilisearch plans native MCP support in the engine, this server is the officially maintained bridge until then.
Tools & capabilities (23)
get-connection-settingsView current Meilisearch URL and API key status.
update-connection-settingsSwitch to a different Meilisearch instance or key at runtime.
create-indexCreate a new index with optional primary key.
list-indexesList all available indexes.
delete-indexDelete an index and all its documents.
get-index-metricsDetailed metrics for a specific index.
get-documentsRetrieve documents with pagination.
add-documentsAdd or update documents in an index.
searchSearch one or multiple indices with filtering and sorting.
get-settingsView current index settings.
update-settingsUpdate ranking, faceting, and other index settings.
get-keysList all API keys.
create-keyCreate an API key with specific permissions.
delete-keyDelete an API key.
get-taskGet information about a specific task.
get-tasksList tasks with optional filters.
cancel-tasksCancel pending or enqueued tasks.
delete-tasksDelete completed tasks.
health-checkBasic health check.
get-health-statusComprehensive health status.
get-versionMeilisearch version information.
get-statsDatabase statistics.
get-system-infoSystem-level information.
What this server can do
Meilisearch MCP Server provides tools for these capabilities — tap one to see every MCP server that does the same:
When to use it
- Prototype a search feature: create an index, load sample docs, and test queries in one chat
- Tune relevance by iterating on ranking rules and searchable attributes conversationally
- Run hybrid keyword + semantic searches across multiple indices
- Monitor indexing tasks and instance health during data imports
Quick setup
- 1Start Meilisearch: docker run -d -p 7700:7700 getmeili/meilisearch:v1.28
- 2Install the server: pip install meilisearch-mcp (or use uvx -n meilisearch-mcp directly)
- 3Add {"command": "uvx", "args": ["-n", "meilisearch-mcp"]} under mcpServers in claude_desktop_config.json
- 4Set MEILI_HTTP_ADDR and MEILI_MASTER_KEY env vars if not using the defaults
- 5Test with 'create an index called products with id as primary key'
Security notes
With the master key configured, the agent can delete indices, rewrite settings, and mint API keys — use a scoped key instead of the master key where possible. The README itself warns that switching hosts/keys via chat tools is a development convenience and not production-grade practice.
Meilisearch MCP Server FAQ
Is meilisearch-mcp free and official?
Yes on both: it is MIT-licensed and maintained in the meilisearch GitHub organization. You still need a Meilisearch instance — self-hosted is free, Meilisearch Cloud is paid.
Does it work with clients other than Claude?
Yes. It is a standard stdio MCP server and the README explicitly targets any MCP-compatible client, including OpenAI agents. There is also a Docker image (getmeili/meilisearch-mcp) for n8n-style containerized setups.
Can it do vector/semantic search?
Yes — the search tooling covers Meilisearch's semantic vector search and hybrid keyword+semantic mode, alongside classic filtered and faceted search. You configure embedders on the Meilisearch side and query through the same search tool.
Alternatives to Meilisearch MCP Server
Compare all alternatives →
Official MCP reference server that fetches a URL and returns its content as clean Markdown, with chunking.
Official Firecrawl MCP server — scrape, crawl, map, search, and structured extraction for any LLM client.
Exa's neural web search and crawling MCP server — runs locally via npx or as a hosted remote endpoint.
Compare Meilisearch MCP Server with: