Execute LangChain, AutoGen, and CrewAI tools safely within our Ephemeral Deterministic Sandboxes.
Bridge standard LangChain agents into containerized bash execution.
Spin up multi-container orchestrations for conversing agent swarms.
Offload heavy Python execution tasks from your CrewAI processes.
Giving an LLM access to a `PythonREPLTool` or `TerminalTool` on your host machine is a massive security risk. Acadify provides a drop-in replacement LangChain Tool that routes all code execution to our ephemeral Kubernetes pods.
from langchain.agents import AgentExecutor, create_react_agent
from acadify.integrations.langchain import AcadifySandboxTool
# Initialize the secure sandbox tool
sandbox_tool = AcadifySandboxTool(
api_key="aca_live_xyz123",
environment="python:3.11",
timeout_seconds=300
)
tools = [sandbox_tool]
# The LLM can now write python scripts and execute them securely.
agent = create_react_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools)
agent_executor.invoke({"input": "Write a script to calculate the first 10,000 prime numbers and benchmark the CPU time."})Microsoft's AutoGen framework allows multiple agents to converse and collaborate. For complex tasks like full-stack web development, you can assign different container environments to different agents.
By registering different Acadify sandboxes as execution nodes, your "Backend Agent" can execute code in a Node.js pod, while your "Data Scientist Agent" executes code in a Jupyter/PyTorch pod. They communicate via AutoGen, but their execution environments remain isolated and secure.