MCP
Model Context Protocol (MCP) is an open protocol standard proposed by Anthropic to standardize the way large language models (LLMs) interact with external tools, data sources, and services regarding contextual information. MCP defines how a model requests, receives, and utilizes external context, supporting capabilities such as tool invocation and resource access. It enables AI applications to integrate diverse external systems in a standardized manner, enhancing model extensibility and interoperability.
Core Facts
When connecting to multiple MCP servers simultaneously, tool descriptions alone can consume thousands or even tens of thousands of tokens.
MCP defines three core primitives: Tools (tool invocation), Resources (resource reading), and Prompts (prompt templates).
MCP aims to address the fragmentation problem in connecting large language models with external data sources and tools.
Timeline (last 90 days)
权限的粒度控制会成为验证 MCP 类工具的核心竞争力。
In the MCP architecture, the timeout timer runs on the client side, and the server cannot relax the client's timeout limit through parameters.
Within the MCP host, the Agent typically does not hold an API key, so performing retrieval directly with the vendor REST API inside the Agent yields a 401, while reading through the tool surface works normally.
Unsloth local models can be used in VS Code Copilot Chat via `unsloth start vscode`; the coding Agent can use Unsloth via MCP with an account API key to perform model loading, generation, training, and export.
herdr's design philosophy is to keep a thin terminal layer, not build every feature into the core, but let agents extend themselves through generic interfaces such as the CLI and MCP.
The same Leafmelt key can be used as a remote MCP server, letting an Agent read documents directly during a conversation.
Jnvoy's MCP security module additionally detects tool poisoning and secret leakage in tool responses.
v4.1.24 fixed the issue where an MCP server or hook exited before reading all input, causing the extension host to crash after about 3 seconds.
Developer selection decision rule: if the capability is ultimately called by code written by a human, using the API directly is more straightforward; if the model is to decide on its own when to call it, then it is worth wrapping another layer of MCP.
Azure AI Language can serve as an MCP Server, invoked through the Tool mechanism of Microsoft Foundry Agent, providing NLP capabilities such as language detection, entity extraction, and PII identification.
40 more timeline events
All Facts (20)
When connecting to multiple MCP servers simultaneously, tool descriptions alone can consume thousands or even tens of thousands of tokens.
90%VerifiedMCP defines three core primitives: Tools (tool invocation), Resources (resource reading), and Prompts (prompt templates).
90%VerifiedMCP aims to address the fragmentation problem in connecting large language models with external data sources and tools.
90%VerifiedMCP采用客户端-服务器架构，基于标准化的JSON-RPC通信协议
90%VerifiedEach tool in MCP has a strict JSON Schema description, including the name, functional description, and parameter specifications
90%VerifiedAnthropic launched MCP (Model Context Protocol) at the end of 2024, an open standard designed to standardize the way AI models interact with external tools and data sources.
90%VerifiedBefore MCP emerged, every AI application needed to write separate integration code for each tool, resulting in extremely high maintenance costs.
90%VerifiedThe design philosophy of MCP draws on LSP (Language Server Protocol), aiming to unify the way context is passed between AI models and external systems.
90%VerifiedThe design of MCP was inspired by the Language Server Protocol (LSP) proposed by Microsoft.
90%VerifiedMCP (Model Context Protocol) is an open standard protocol released by Anthropic in late 2024.
90%VerifiedAnthropic proposed the MCP (Model Context Protocol) in late 2024, attempting to establish a unified standard for communication between agents.
90%VerifiedMCP Server 可以暴露工具（Tools）、资源（Resources）和提示词模板（Prompts）三类能力
90%VerifiedMCP standardizes the integration complexity between AI applications and external tools from M×N to M+N.
90%VerifiedWorkBuddy connects to external systems such as Tencent Docs, WeCom, and databases via MCP (Model Context Protocol).
80%VerifiedThe core design philosophy of DeepSeek Harness is to treat everything as a plugin, including tools, skills, and drawing functionality.
80%VerifiedIndirect prompt injection attacks plant hidden instructions within normal business data to induce Agents into executing unauthorized operations, representing a security risk at the semantic level rather than the code level.
80%VerifiedIn MCP, Tools are invoked at the model's discretion, Prompts are triggered by the user, and Resources are controlled by the application.
80%VerifiedMCP's design was inspired by the Language Server Protocol (LSP), which simplifies the M×N integration problem between editors and programming languages to linear complexity of M+N.
80%VerifiedMCP协议支持两种传输方式：stdio（标准输入输出）和HTTP SSE（Server-Sent Events）
80%Verified在MCP出现之前，AI调用外部工具主要依赖OpenAI的Function Calling机制或各框架自定义的插件系统
80%