Specification and documentation for the Model Context Protocol
About Model Context Protocol
The Model Context Protocol, or MCP, is an open standard that Anthropic released in late November 2024 for connecting AI assistants to the systems where data and tools live, such as content repositories, business tools and development environments.
It uses a client-server design. Hosts are the AI applications that start connections, clients are connectors inside the host, and servers provide context and capabilities. Servers expose resources, tools and prompts, and can ask the client for model completions through sampling. Messages use JSON-RPC 2.0, with standard input and output or HTTP as the specified transports, and the client keeps control over permissions.
The specification is MIT licensed and comes with reference SDKs. Launch coverage mentioned TypeScript and Python, and later sources list more languages, so check the official documentation. A repository of prebuilt servers covered services such as Google Drive, Slack, GitHub, Postgres and Stripe, and many other tools in this directory now speak MCP.
Key features
- Open standard for connecting AI assistants to tools and data
- Client-server design over JSON-RPC 2.0
- Servers expose resources, tools and prompts
- Sampling lets servers request model completions
- Client controls permissions
- MIT-licensed specification with SDKs
- Repository of prebuilt servers
Links
All content and agents from Model Context Protocol
Frequently asked questions about Model Context Protocol
- What is Model Context Protocol?
- Specification and documentation for the Model Context Protocol. It is listed on botshub under agent infrastructure and tools.
- Is Model Context Protocol free or open source?
- Model Context Protocol is tagged as open source on botshub. Check the official site for current terms.
- What can Model Context Protocol do?
- Key features include open standard for connecting AI assistants to tools and data, client-server design over JSON-RPC 2.0, servers expose resources, tools and prompts, sampling lets servers request model completions and client controls permissions.
- What are alternatives to Model Context Protocol?
- Other agent infrastructure and tools on botshub include TypeChat, GitHub MCP Server, LlamaIndex Tools and Cognee. See all alternatives to Model Context Protocol.
