Examples

Code Mode in Pydantic AI

Pydantic AI runs Monty behind CodeModeToolset. Instead of making sequential tool calls, the model writes Python that calls your tools as functions, and Monty executes it: one round trip for a task that would otherwise take three.

import asyncio
import json

import logfire
from httpx import AsyncClient
from pydantic_ai import Agent, RunContext
from pydantic_ai.toolsets.code_mode import CodeModeToolset
from pydantic_ai.toolsets.function import FunctionToolset
from typing_extensions import TypedDict

logfire.configure()
logfire.instrument_pydantic_ai()


class LatLng(TypedDict):
    lat: float
    lng: float


weather_toolset: FunctionToolset[AsyncClient] = FunctionToolset()


@weather_toolset.tool
async def get_lat_lng(
    ctx: RunContext[AsyncClient], location_description: str
) -> LatLng:
    """Get the latitude and longitude of a location."""
    # NOTE: the response here will be random, and is not related to the location description.
    r = await ctx.deps.get(
        'https://demo-endpoints.pydantic.workers.dev/latlng',
        params={'location': location_description},
    )
    r.raise_for_status()
    return json.loads(r.content)


@weather_toolset.tool
async def get_temp(ctx: RunContext[AsyncClient], lat: float, lng: float) -> float:
    """Get the temp at a location."""
    # NOTE: the responses here will be random, and are not related to the lat and lng.
    r = await ctx.deps.get(
        'https://demo-endpoints.pydantic.workers.dev/number',
        params={'min': 10, 'max': 30},
    )
    r.raise_for_status()
    return float(r.text)


@weather_toolset.tool
async def get_weather_description(
    ctx: RunContext[AsyncClient], lat: float, lng: float
) -> str:
    """Get the weather description at a location."""
    # NOTE: the responses here will be random, and are not related to the lat and lng.
    r = await ctx.deps.get(
        'https://demo-endpoints.pydantic.workers.dev/weather',
        params={'lat': lat, 'lng': lng},
    )
    r.raise_for_status()
    return r.text


agent = Agent(
    'gateway/anthropic:claude-sonnet-4-5',
    toolsets=[CodeModeToolset(weather_toolset)],
    deps_type=AsyncClient,
)


async def main():
    async with AsyncClient() as client:
        await agent.run('Compare the weather of London, Paris, and Tokyo.', deps=client)


if __name__ == '__main__':
    asyncio.run(main())

Swap CodeModeToolset(weather_toolset) for weather_toolset to see the same task done with ordinary tool calls.

Worked examples in the repository

Each directory under examples/ is runnable after make dev-py; its README has the command.

• sql_playground: customer purchase data in CSV joined with tweets in JSON, with sentiment analysis called in a loop from the sandbox. With JSON tool calling the 50+ per-tweet results would flood the context window; in Monty they stay inside the sandbox and only the aggregate comes out. Also shows file sandboxing via the os callback and type checking against a stub file.
• expense_analysis: Anthropic’s programmatic tool calling cookbook example, run on Monty.
• web_scraper: Playwright and BeautifulSoup exposed to the sandbox as host objects so the model can extract prices from model labs’ websites; example_code.py is the code Claude Sonnet 4.5 wrote for it.
• classes: one short file per behaviour of host objects, in Python and TypeScript: explicit policies, lazy attributes, sandbox-side copies, convert_value hooks, constructing host classes from the sandbox, and round-tripping sandbox-defined classes.