langchain-disposable

Give your LangChain agent a code execution tool that runs in an isolated container — no network, read-only filesystem, 30-second timeout. Every run is disposable.

When to use it

Use this package if you already use LangChain and want:

• a LangChain tool for remote execution
• API-key-authenticated execution through Disposable Exec
• the standard Disposable Exec flow behind a LangChain-facing interface

If you just need the raw client, use disposable-exec.

Install

pip install langchain-disposable disposable-exec

For the agent example, also install the LangChain packages and model provider package you already use.

Environment variables

export DISPOSABLE_EXEC_BASE_URL="https://your-api-host"
export DISPOSABLE_EXEC_API_KEY="de_xxx"

Minimal tool example

import os

from langchain_disposable import DisposableExecTool

tool = DisposableExecTool(
    api_key=os.environ["DISPOSABLE_EXEC_API_KEY"],
    base_url=os.environ["DISPOSABLE_EXEC_BASE_URL"],
)

result = tool.invoke({"script": "print(2 + 2)"})
print(result)

Minimal agent example

import os

from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI

from langchain_disposable import DisposableExecTool

tool = DisposableExecTool(
    api_key=os.environ["DISPOSABLE_EXEC_API_KEY"],
    base_url=os.environ["DISPOSABLE_EXEC_BASE_URL"],
)

prompt = ChatPromptTemplate.from_messages(
    [
        ("system", "Use the execution tool when Python is useful."),
        ("human", "{input}"),
        ("placeholder", "{agent_scratchpad}"),
    ]
)

agent = create_tool_calling_agent(ChatOpenAI(model="gpt-4o-mini"), [tool], prompt)
executor = AgentExecutor(agent=agent, tools=[tool], verbose=True)

print(executor.invoke({"input": "Use Python to compute the mean of [2, 4, 6, 8]."}))

What the agent gets back

The tool returns a JSON string with:

• stdout — captured standard output
• stderr — captured standard error
• exit_code — process exit code (0 = success)
• failure_reason — structured reason if failed: timeout, exit_nonzero, sandbox_error, platform_error, orphaned
• duration — execution time in seconds

Your agent can use failure_reason to decide whether to retry, rewrite the code, or give up.

Relationship to Disposable Exec

langchain-disposable is a thin LangChain wrapper on top of disposable-exec. Disposable Exec provides the API key, /me, /run, /status, and /result flow. This package makes that flow easy to use from LangChain.