Maintained by deepset

Integration: E2B

Use E2B cloud sandboxes as tools in a Haystack Agent to run bash commands and read, write, and list files in an isolated Linux environment

Authors

deepset

GitHub Repo PyPI Package

Overview
Installation
Usage
License

Overview

E2B provides secure, isolated cloud sandboxes that let LLM-powered agents execute arbitrary code and shell commands without touching the host machine. The e2b-haystack integration wraps the E2B SDK as a set of Haystack Tools that an Agent can invoke to run bash commands and manage files inside a shared sandbox.

All tools provided by the integration operate on the same live E2BSandbox, so files written by one tool call are immediately available to the next — the agent can, for example, write a Python script with write_file, execute it with run_bash_command, and read the result back with read_file.

Installation

pip install e2b-haystack

Set your E2B API key as an environment variable:

export E2B_API_KEY="your-e2b-api-key"

Usage

Components

This integration introduces the following tools and a toolset wrapper:

RunBashCommandTool: Runs a bash command inside the sandbox and returns stdout, stderr, and exit code.
ReadFileTool: Reads a file from the sandbox filesystem.
WriteFileTool: Writes content to a file in the sandbox filesystem.
ListDirectoryTool: Lists the contents of a directory inside the sandbox.
E2BToolset: A bundle of all four tools sharing a single E2BSandbox, ready to plug into an Agent.

Use with a Haystack Agent

The E2BToolset enables you to give an Agent access to a sandbox:

from haystack.components.agents import Agent
from haystack.components.generators.chat import OpenAIChatGenerator
from haystack.dataclasses import ChatMessage

from haystack_integrations.tools.e2b import E2BToolset

agent = Agent(
    chat_generator=OpenAIChatGenerator(model="gpt-4o-mini"),
    tools=E2BToolset(sandbox_template="base", timeout=120),
    system_prompt=(
        "You are a helpful coding assistant with access to a live Linux sandbox. "
        "Use the available tools freely to explore, write files, and run commands."
    ),
    max_agent_steps=10,
)

result = agent.run(
    messages=[
        ChatMessage.from_user(
            "Write a Python script to /tmp/primes.py that prints all prime numbers "
            "up to 50, run it, then read the file back so I can see both the script "
            "and its output."
        )
    ]
)

print(result["last_message"].text)

If you prefer to assemble the tools yourself, instantiate a single E2BSandbox and pass it to each tool so they share the same environment:

from haystack.components.agents import Agent
from haystack.components.generators.chat import OpenAIChatGenerator
from haystack.dataclasses import ChatMessage

from haystack_integrations.tools.e2b import (
    E2BSandbox,
    ListDirectoryTool,
    ReadFileTool,
    RunBashCommandTool,
    WriteFileTool,
)

sandbox = E2BSandbox()
tools = [
    RunBashCommandTool(sandbox=sandbox),
    ReadFileTool(sandbox=sandbox),
    WriteFileTool(sandbox=sandbox),
    ListDirectoryTool(sandbox=sandbox),
]

agent = Agent(
    chat_generator=OpenAIChatGenerator(model="gpt-4o-mini"),
    tools=tools,
    max_agent_steps=15,
)

result = agent.run(
    messages=[ChatMessage.from_user("Generate the first 10 Fibonacci numbers using bash.")]
)
print(result["last_message"].text)

License

e2b-haystack is distributed under the terms of the Apache-2.0 license.