X's own hosted MCP server: full-archive post search, user and timeline lookup, trends, bookmarks and article drafting, under your account's permissions.
Typesense CloudMCP Server
Summary
Typesense's official remote MCP server — search collections, index documents, and create, resize or clone the Typesense Cloud clusters behind them, over OAuth with per-agent permission presets.
Features
- Remote streamable-HTTP endpoint at cloud.typesense.org/mcp/v1 — nothing to install
- Eight tools spanning search, document indexing, cluster management and analytics
- Vector and hybrid search over Typesense collections
- Create, resize, rename, clone and terminate clusters from the agent
- Version-matched documentation search so answers fit the cluster you run
- Five OAuth permission presets, capped by your own role on team accounts
- Rate limits of 300 data calls/min and 30 cluster actions/min
Installation
Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.
Description
Typesense Cloud's official MCP server does something most search integrations stop short of: as well as querying an index, it lets an agent operate the infrastructure under it. The same connection that runs a hybrid vector search can spin up a cluster, resize it, clone it or tear it down.
It is a remote server at https://cloud.typesense.org/mcp/v1 over streamable HTTP, so there is nothing to install.
The eight tools
whoami— reports which account the agent is acting for and what it is allowed to dosearch_typesense_docs— searches the documentation matched to your cluster's version, so the answer is not for a release you are not runningsearch_documents— collection search, including vector and hybrid queriesget_clusters— lists clusters with configuration, metrics and billingmanage_cluster— create, resize, rename, clone and terminatedocuments— index, update, fetch, delete, import and exportread_cluster_data/change_cluster_data— collections, aliases, synonyms, curation and analytics
Permissions are the interesting part
Authentication runs through a browser consent screen where you sign in, pick which account the agent works on, and choose its permissions before connecting. Five presets bound what it can reach — Search only, Search and curate, Manage clusters, Build (the default) and Everything — and individual permissions can be toggled within them. On team accounts the agent can never exceed your own role, so a connector cannot quietly escalate. Headless environments skip the browser and pass a Cluster Management API key as a bearer token instead.
Limits worth knowing before you wire it up
Data operations are capped at 300 calls per minute and cluster management at 30 actions per minute. Inline imports are limited to 1,000 documents or 10 MB; anything larger is handed back as a temporary curl command — which means browser-based assistants, unable to run shell commands, cannot complete large transfers. Free-tier clusters (0.5 GB RAM, 2 vCPU) need no payment method; going beyond that requires one or prepaid credits.
A good fit for teams building typo-tolerant site search who want an agent to handle both the index and the cluster, and for anyone prototyping a search schema without leaving the editor.
Related MCP Servers
Official MCP server for Meilisearch — create and configure indexes, add documents, run hybrid and semantic searches, and manage API keys and tasks through conversation.
Apify's official MCP server — turn thousands of ready-made scrapers into tools your assistant can call.
Browserbase's official MCP server — cloud browser automation with Stagehand's natural-language actions.