Monty Python REPL¶

SuperQode can expose an optional python_repl tool backed by pydantic-monty. Monty is a small Python interpreter written in Rust for running agent-written snippets with tighter controls than direct host Python execution.

This is not a model provider. It is a tool the coding agent can use for quick calculations, parsing, small transformations, and controlled interpreter-style work during a session.

Install¶

Monty support is optional:

uv sync --extra monty

For installed packages:

pip install 'superqode[monty]'

Verify the dependency:

superqode providers monty check

Run a smoke test:

superqode providers monty smoke

Tool Availability¶

When pydantic-monty is installed, SuperQode registers python_repl in the standard and full tool profiles.

If Monty is not installed, SuperQode does not expose the tool to the model. This keeps the default install small and avoids failing sessions on missing optional dependencies.

Behavior¶

python_repl runs a Python snippet in a fresh Monty sandbox and returns the captured print() output plus the value of the final expression.

Behavior	Default
Isolation	Each call runs in a fresh sandbox - no state carries between calls
Host filesystem	No access (`open` is not defined in the sandbox)
Network	No access
Environment / stdlib	No third-party imports; only a minimal built-in subset
Execution limit	Duration and memory caps
Output limit	Truncated before returning to the model

The tool parameters are:

Parameter	Default	Description
`code`	-	Python snippet to execute. Use `print()` to emit output.
`type_check`	`false`	Type-check the snippet before running it
`max_duration_secs`	`2.0`	Maximum execution time in seconds
`max_memory`	`32 MiB`	Maximum Monty heap memory in bytes

Sandbox Isolation¶

Monty has no access to the host. The filesystem, environment variables, and network are all unavailable inside the sandbox - for example open(...) raises NameError and import socket raises ModuleNotFoundError. This is what makes it safe to run model-written snippets in-process without a container.

If a task needs to read or write real project files, use the read_file, write_file, or edit_file tools (which run under the permission system), or bash for shell access.

Note: Monty is experimental and implements a subset of Python. Snippets that need third-party packages (NumPy, requests, ...), the full standard library, or the project's own modules should use bash with the project's Python interpreter instead.

When To Use¶

Good uses:

Small calculations
Parsing structured text
Trying a tiny algorithm
Transforming JSON or tabular data
Testing a small pure-Python expression

Use bash or project test commands instead when the task needs:

Third-party Python packages from the project environment
Full CPython compatibility
Running the project's real test suite
Shell commands, package managers, or external programs

Troubleshooting¶

If python_repl is missing:

superqode providers monty check

If the check reports that pydantic-monty is not installed, install the optional extra and restart the SuperQode session:

uv sync --extra monty

If code cannot read files, call the network, or import third-party packages, that is expected - Monty is fully isolated from the host by design. Use read_file/bash for those tasks instead.