Skip to content

Commit f104e8b

Browse files
committed
MCP overview
1 parent 9070308 commit f104e8b

File tree

2 files changed

+38
-1
lines changed

2 files changed

+38
-1
lines changed

docs/use-cases/AI_ML/MCP/index.md

Lines changed: 38 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -4,10 +4,47 @@ sidebar_label: 'MCP'
44
title: 'MCP guides'
55
pagination_prev: null
66
pagination_next: null
7-
description: 'This page contains a landing page with table of contentsfor MCP guides.'
7+
description: 'This page provides an intro to Model Context Protocol (MCP) and has a table of contents for MCP guides.'
88
keywords: ['ClickHouse', 'MCP']
99
show_related_blogs: true
1010
---
1111

12+
import Image from '@theme/IdealImage';
13+
import overview from '@site/static/images/use-cases/AI_ML/MCP/mcp_overview.png';
14+
15+
[Model Context Protocol](https://www.anthropic.com/news/model-context-protocol) (MCP) is a standard developed by Anthropic that enables AI models to seamlessly integrate with external systems. This protocol allows AI assistants to connect to data sources, APIs, databases, and more in a secure and standardized way.
16+
17+
MCP creates a universal interface between AI models and various services, eliminating the need for custom tool implementations for each integration.
18+
You can think of it as a universal API standard designed specifically for AI systems.
19+
20+
The key advantage of MCP is that AI libraries only need to implement support for the protocol once.
21+
After that, all MCP-compatible services become immediately accessible, saving AI library maintainers a lot of time.
22+
23+
## What is MCP's architecture? {#mcp-architecture}
24+
25+
MCP follows a client-server architecture:
26+
27+
* Clients (like Claude Desktop, Cursor, or VS Code) establish connections with MCP servers. You can see a collection of clients in the [awesome-mcp-clients](https://github.com/punkpeye/awesome-mcp-clients?tab=readme-ov-file#windsurf) GitHub repository.
28+
* Servers expose tools and capabilities through standardized interfaces. You can see a collection of servers in the [awesome-mcp-servers](https://github.com/punkpeye/awesome-mcp-servers) GitHub repository.
29+
* AI models can then use these tools to access external data and functionality when needed
30+
31+
A diagram showing the architecture is below:
32+
33+
<Image img={overview} alt="MCP overview" size="lg"/>
34+
35+
The Claude Desktop application acts as a client that can connect to various MCP servers, enabling Claude to interact with external services like Google Drive, web APIs, Google Maps, and databases—including ClickHouse.
36+
37+
## Does ClickHouse have an MCP Server? {#clickhouse-mcp-server}
38+
39+
It does!
40+
The [ClickHouse MCP Server](https://github.com/ClickHouse/mcp-clickhouse) offers the following tools:
41+
42+
* `run_select_query` - Execute SQL queries on your ClickHouse cluster.
43+
* `list_databases` - List all databases on your ClickHouse cluster.
44+
* `list_tables` - List all tables in a database.
45+
46+
47+
## Guides for using the ClickHouse MCP Server {#clickhouse-mcp-server-guides}
48+
1249
<!--AUTOGENERATED_START-->
1350
<!--AUTOGENERATED_END-->
66.9 KB
Loading

0 commit comments

Comments
 (0)