Skip to content

Memory & Storage Tools

Notes — add_note & read_notes

Simple file-based note taking. Demonstrates file I/O in MCP tools.

@mcp.tool()
def add_note(content: str) -> str: ...

@mcp.tool()
def read_notes() -> str: ...

Example prompts

Add a note: "MCP uses stdio transport by default"
Save this as a note: "Always unset VIRTUAL_ENV before running uv"
Read all my notes
What notes have I saved?

Member Database — add_person_to_member_database

Demonstrates structured Pydantic input — Claude constructs a typed Person object from natural language.

class Person(BaseModel):
    first_name: str
    last_name: str
    years_of_experience: int
    previous_addresses: List[str]

@mcp.tool()
def add_person_to_member_database(person: Person) -> str: ...

Example prompts

Add Jane Smith to the member database — she has 7 years of experience
  and previously lived at 12 Oak Street and 45 Maple Avenue.

Log a new member: John Doe, 3 years experience, no previous addresses.

Add me to the database: first name Alex, last name Johnson,
  12 years experience, previously at 100 Main St, New York.

Why this matters

This shows how MCP + Pydantic lets Claude turn free-form natural language into validated, structured data — no parsing code needed.

Vector Memory — save_memory & search_memory

Persists knowledge to an OpenAI vector store and retrieves it semantically. Requires OPENAI_API_KEY.

@mcp.tool()
def save_memory(memory: str) -> dict: ...

@mcp.tool()
def search_memory(query: str) -> dict: ...

Example prompts

Save this to memory: "The MCP ToolHub project uses FastMCP and uv for packaging"

Remember that: "Perplexity sonar-pro is the best model for web search in this server"

Search my memories for anything about MCP configuration

What do I know about web search setup?

Requires OpenAI API key

Set OPENAI_API_KEY in your .env file before using memory tools.