What is a remote MCP server?
A remote MCP server runs over HTTP/SSE instead of a local process — you point your client at a URL instead of installing and running anything on your own machine.
Official hosted MCP endpoints from SaaS providers — connect directly, no install required.
Airtable
Bring your structured data to Claude
Google Cloud BigQuery
BigQuery: Advanced analytical insights for agents
MotherDuck
DuckDB cloud warehouse, SQL
Supabase
Projects, database, docs
PlanetScale
Authenticated access to your Postgres and MySQL DBs
Neon
Postgres projects, branches, SQL, docs
pg-aiguide
Search pg and Tiger docs, learn database skills
Common questions about hosted and remote MCP servers
A remote MCP server runs over HTTP/SSE instead of a local process — you point your client at a URL instead of installing and running anything on your own machine.
A local server runs on your machine and needs a runtime like Node.js or Python. A remote server is hosted by a third party — you just need a URL (and sometimes an API key) to connect, which is simpler to set up but depends on the provider staying online.
Your data passes through the provider's infrastructure, so stick to servers run by official or well-known teams that document how they handle data, and grant sensitive permissions carefully.
Add the server URL in a client that supports remote MCP, then configure OAuth, an API key, or request headers as required by the provider. Streamable HTTP, SSE, and authentication support vary by client, so follow both the server page and client documentation.
Yes. Calls depend on network access, provider uptime, and rate limits. For important workflows, check the provider's status page, timeout and retry behavior, quotas, and fallback options.
Yes, when the client supports the server's remote transport and authentication method. Newer clients commonly support Streamable HTTP, while older versions may only support SSE or local stdio, so verify version compatibility first.