langchain-leap0 0.2.0

Leap0 sandbox integration for Deep Agents

langchain-leap0

Leap0 integration for LangChain Deep Agents: a BaseSandbox backend that runs shell commands and file transfers inside a Leap0 sandbox.

Leap0 is enterprise-grade cloud sandboxes for AI agents. Launch isolated sandboxes in ~200ms. Give every agent its own compute, filesystem, and network boundary while your agents run safely.

Quick install

pip install langchain-leap0

Set your API key (see Leap0 for account and key management):

export LEAP0_API_KEY="your-key"

Minimal usage with the [leap0](https://pypi.org/project/leap0/) SDK and Leap0Sandbox:

from leap0 import Leap0Client

from langchain_leap0 import Leap0Sandbox

client = Leap0Client()
sandbox = client.sandboxes.create()
backend = Leap0Sandbox(client=client, sandbox=sandbox)

try:
    result = backend.execute("echo 'Hello LangChain from Leap0.dev'")
    print(result.output)
finally:
    client.sandboxes.delete(sandbox)
    client.close()

Examples

Runnable scripts for Deep Agents sandboxes. See [examples/README.md](examples/README.md) for environment variables and how they map to LangChain **SandboxIntegrationTests**.

Script	Summary
examples/basic_sandbox.py	Minimal: create a sandbox, Leap0Sandbox.execute(), teardown.
examples/deep_agent_sandbox.py	create_deep_agent(..., backend=Leap0Sandbox(...)) — create and run a script; prints a short execute summary.

From the repo root:

uv run python examples/basic_sandbox.py
uv run python examples/deep_agent_sandbox.py

Developing from source

From this directory (alongside leap0-python in the Leap0 workspace). Uses pyproject.toml dependencies; run:

cd langchain-leap0
uv sync --all-groups

Running tests

From the package root:

# Unit tests
make test

# LangChain standard sandbox integration tests
make integration_test

Integration tests require:

export LEAP0_API_KEY="your-key"

Documentation

Full documentation is available at leap0.dev/docs.

License

Apache License 2.0. See LICENSE for details.