Typesense Cloud MCP. Diagnose search latency and cluster status instantly.

Look, you're running a search cluster. It's complex, right? You don't want to be logged into some terminal just poking around with curl commands trying to figure out why latency spiked last Tuesday. This server lets your agent talk directly to the Typesense backend. Instead of manually checking things, your AI client does it for you.

When you connect this endpoint, you get full diagnostic control over a live search cluster. You're not just getting simple status lights; you're pulling quantitative data so you know exactly what's happening under the hood.

Checking Core Operational Status

To see if everything's running right, your agent uses get_cluster_health. This tool checks all nodes in the cluster to confirm they're online and responding normally. It tells you immediately if any part of the system is down or struggling with basic connectivity issues.

If you need deeper performance intel, you run get_cluster_metrics. This pulls specific numbers: search latency—you get both the median time and the 99th percentile—along with CPU usage reports and logs detailing overall resource consumption. You'll see patterns in how much work the cluster is handling.

Managing Data Structure and Access Points

You gotta know what data you’re searching, and who has keys to it. To get a list of every searchable collection configured on the cluster, call list_collections. This gives you the names of all your actual data sources.

When you need to audit how users access those collections, use list_collection_aliases. This tool lists virtual aliases; these are clean public-facing names that map back to your real, internal data collections. It keeps your API management tidy.

For security, always check the keys. Run list_api_keys to pull a complete list of every active and configured API key for the whole cluster. You'll know exactly who has access and what they can do with it.

Running Complex Searches

The most powerful thing here is running queries efficiently. Instead of sending one search request at a time, you use execute_multi_search. This tool lets your agent send a JSON array containing multiple independent search requests in a single API call. You can query several different collections simultaneously and get all the raw JSON results back without having to juggle ten separate calls.

Think of it like this: Need to check inventory across five regions, and also run a trending topic search on articles, all at once? You bundle those into one request using execute_multi_search. The agent handles the complex payload structure, so you just get clean results. It’s about batching work—getting everything done in one go.

This gives your AI client native control over diagnosis and querying. You don't need to learn a bunch of command-line syntax; you just tell your agent what job needs doing, whether it's checking the 99th percentile latency or running five different searches across three collections. The agent executes the right tool call against the Typesense backend.