toolschema documentation¶

toolschema turns typed Python functions into JSON Schema for AI agent tools — once, for every provider.

Your function  →  schema(fn)  →  ToolDefinition  →  OpenAI / MCP / LangChain / …

Who is this for?¶

You are…	Start here
New to toolschema	Installation → Quick start
Building an MCP server	MCP tutorial → Scaffolding
Using LangChain / Agents SDK	Integrations
Looking up a function	API reference
Understanding design	Architecture

What toolschema does¶

Reads your function's type hints, defaults, and docstring
Builds a canonical JSON Schema 2020-12 (ToolDefinition)
Exports provider-specific JSON (OpenAI, Anthropic, Gemini, MCP)
Integrates with FastMCP, LangChain, OpenAI Agents, Pydantic AI via thin adapters

What toolschema does not do¶

Run agents or orchestrate LLM calls
Replace Pydantic for domain modeling
Handle MCP transport (use FastMCP for that)

Install¶

pip install toolschema
pip install toolschema[fastmcp]   # MCP servers
pip install toolschema[all]       # everything

See Installation for all extras.

Minimal example¶

from toolschema import tool, schema

@tool
def greet(name: str) -> str:
    """Say hello."""
    return f"Hello, {name}!"

definition = schema(greet)
print(definition.to_openai())
print(definition.to_mcp())

Package layout¶

toolschema/
├── tool, schema, Field          # public API
├── ToolDefinition               # IR (intermediate representation)
├── adapters/                    # OpenAI, Anthropic, Gemini, MCP
├── integrations/                # FastMCP, LangChain, Agents, Pydantic AI
├── cli.py                       # inspect, diff, export, init
└── templates/                   # MCP server scaffold

Version¶

Current release: 1.0.0 (see PyPI).