What is MCP Prompt?
MCP Prompt is a reusable prompt template exposed by an MCP Server so MCP clients and hosts can present consistent task-specific instructions, arguments, and interaction patterns.
Quick Facts
| Specification | Official Specification |
|---|
How It Works
An MCP Prompt lets a server package domain-specific guidance as a reusable prompt template. It is useful when a tool or resource set needs a recommended way to be used, such as code review, log analysis, database investigation, or document summarization. Prompts are not executable tools, and they are not a substitute for application policy. They are structured guidance that a host may present to the user or model with explicit arguments.
Key Characteristics
- Reusable instruction asset: packages task-specific guidance in a server-provided template
- Argument-aware: can declare parameters that the host or user fills before use
- Domain-specific: often reflects how a server's tools and resources should be combined
- Non-executable: provides guidance rather than directly performing an operation
- Governance-sensitive: should not smuggle hidden instructions that override host policy
Common Use Cases
- Providing a standard code-review prompt for a repository MCP Server
- Guiding an AI assistant through log investigation with required fields
- Offering a database analysis prompt that references available schema resources
- Standardizing document summarization instructions for a knowledge-base server
- Packaging team-approved workflows for internal AI assistants
Example
Loading code...Frequently Asked Questions
How is an MCP Prompt different from an MCP Tool?
An MCP Prompt provides reusable guidance. An MCP Tool executes an operation. A prompt may tell the model how to use tools, but it does not itself perform the tool action.
Who decides whether an MCP Prompt is used?
The host or user interface decides how prompts are surfaced. A model may use prompt content only after the host includes it in the interaction.
Can MCP Prompts include arguments?
Yes. Prompts can define arguments such as a repository path, issue ID, log query, or target document. This makes them reusable without hardcoding one task instance.
What is a security concern with MCP Prompts?
Prompts should not contain hidden instructions that bypass host policy, ignore user consent, or instruct the model to misuse tools. They should be reviewed like any other model-facing instruction.