> ## Documentation Index
> Fetch the complete documentation index at: https://docs.langchain.com/llms.txt
> Use this file to discover all available pages before exploring further.

# DatabricksSandbox integration

> Integrate with the DatabricksSandbox sandbox backend using LangChain Python.

[Databricks](https://www.databricks.com/) sandboxes are managed, isolated compute environments on the Databricks platform. `DatabricksSandbox` runs [Deep Agents](https://github.com/langchain-ai/deepagents) shell and filesystem tools inside a Databricks sandbox over the [`databricks-sdk`](https://pypi.org/project/databricks-sdk/). See the [Databricks docs](https://docs.databricks.com/) for workspace setup and authentication.

> [!NOTE]
>
> The `DatabricksSandbox` backend requires `databricks-langchain>=0.22.0`.

## Installation

```bash pip theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
pip install 'databricks-langchain[deepagents]>=0.22.0'
```

```bash uv theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
uv add 'databricks-langchain[deepagents]>=0.22.0'
```

## Create a sandbox backend

`DatabricksSandbox` binds to a sandbox that is already running. Provision (or reuse) a sandbox with the Databricks SDK, wait for it to reach `RUNNING`, then wrap it as a [deepagents backend](https://docs.langchain.com/oss/python/deepagents/backends.md). The sandbox lifecycle (create, start, delete) stays in your application code, not the agent.

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
import time

from databricks.sdk import WorkspaceClient
from databricks.sdk.service.sandbox import (
    ComputeSpec,
    Sandbox,
    SandboxSpec,
    SandboxState,
)
from google.protobuf.duration_pb2 import Duration

from databricks_langchain import DatabricksSandbox

client = WorkspaceClient()

sandbox_id = "my-agent-sandbox"
name = f"sandboxes/{sandbox_id}"

client.sandbox.create_sandbox(
    Sandbox(
        spec=SandboxSpec(
            compute=ComputeSpec(inactivity_timeout=Duration(seconds=900))
        )
    ),
    sandbox_id=sandbox_id,
)

# Wait until the sandbox is RUNNING before running commands.
while client.sandbox.get_sandbox(name).status.state != SandboxState.SANDBOX_STATE_RUNNING:
    time.sleep(3)

backend = DatabricksSandbox(client, name)

result = backend.execute("echo hello")
print(result.output)
```

## Use with Deep Agents

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
from deepagents import create_deep_agent
from langchain_anthropic import ChatAnthropic

# `backend` from the previous step, bound to a RUNNING Databricks sandbox.
agent = create_deep_agent(
    model=ChatAnthropic(model="claude-sonnet-4-6"),
    system_prompt="You are a coding assistant with sandbox access.",
    backend=backend,
)

result = agent.invoke(
    {"messages": [{"role": "user", "content": "Write and run a Python script"}]}
)
```

You can use any LangChain chat model, including [`ChatDatabricks`](https://docs.langchain.com/oss/python/integrations/chat/databricks) backed by Databricks Model Serving.

## Cleanup

You are responsible for managing the sandbox lifecycle.
Sandboxes consume resources until deleted, so delete the sandbox when you are done.

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
client.sandbox.delete_sandbox(name)
```

See also: [Sandboxes](https://docs.langchain.com/oss/python/deepagents/sandboxes.md).

***

[Connect these docs](https://docs.langchain.com/use-these-docs.md) to your agent of choice via MCP for real-time answers.

[Edit this page on GitHub](https://github.com/langchain-ai/docs/edit/main/src/oss/python/integrations/sandboxes/databricks.mdx) or [file an issue](https://github.com/langchain-ai/docs/issues/new/choose).