Skip to main content

agentkernel

Python SDK for agentkernel — run AI coding agents in secure, isolated microVMs.

Install

pip install agentkernel-sdk

Requires Python 3.10+.

Quick Start

from agentkernel import AgentKernel

with AgentKernel() as client:
    result = client.run(["echo", "hello"])
    print(result.output)  # "hello\n"

Async

from agentkernel import AsyncAgentKernel

async with AsyncAgentKernel() as client:
    result = await client.run(["echo", "hello"])
    print(result.output)

Sandbox Sessions

with AgentKernel() as client:
    with client.sandbox("test", image="python:3.12-alpine") as sb:
        sb.run(["pip", "install", "numpy"])
        result = sb.run(["python3", "-c", "import numpy; print(numpy.__version__)"])
        print(result.output)
    # sandbox auto-removed

Exec Options

Run commands with a working directory, environment variables, or as root:

with AgentKernel() as client:
    result = client.exec_in_sandbox(
        "my-sandbox",
        ["npm", "start"],
        workdir="/app",
        env=["NODE_ENV=production"],
        sudo=True,
    )

Works on sandbox sessions too:

with client.sandbox("dev") as sb:
    sb.run(["pip", "install", "-r", "requirements.txt"], workdir="/app", sudo=True)

Git Source Cloning

Clone a git repo into the sandbox at creation time:

with AgentKernel() as client:
    client.create_sandbox(
        "my-project",
        image="node:20-alpine",
        source_url="https://github.com/user/repo.git",
        source_ref="main",
    )

Persistent Volumes

Mount volumes that persist across sandbox restarts:

# First create volumes via CLI: agentkernel volume create mydata

with AgentKernel() as client:
    client.create_sandbox(
        "my-project",
        image="python:3.12-alpine",
        volumes=["mydata:/data", "cache:/tmp/cache:ro"],
    )

    # Data in /data persists across sandbox restarts
    client.exec_in_sandbox("my-project", ["sh", "-c", "echo hello > /data/test.txt"])

File Operations

Read, write, and delete files in a sandbox:

with AgentKernel() as client:
    # Write a file
    client.write_file("my-sandbox", "app/main.py", "print('hello')")

    # Read a file
    file = client.read_file("my-sandbox", "app/main.py")
    print(file.content)

    # Delete a file
    client.delete_file("my-sandbox", "app/main.py")

    # Batch write multiple files at once
    client.write_files("my-sandbox", {
        "/app/index.js": "console.log('hi')",
        "/app/package.json": '{"name":"app"}',
    })

Detached Commands

Run long-lived processes in the background and retrieve their output later:

with AgentKernel() as client:
    # Start a background process
    cmd = client.exec_detached("my-sandbox", ["python3", "train.py"])
    print(f"Started: {cmd.id} (pid {cmd.pid})")

    # Check status
    status = client.detached_status("my-sandbox", cmd.id)
    print(status.status)  # "running" | "completed" | "failed"

    # Get logs
    logs = client.detached_logs("my-sandbox", cmd.id)
    print(logs.stdout)

    # Get stderr only
    stderr = client.detached_logs("my-sandbox", cmd.id, stream="stderr")

    # List all detached commands
    all_cmds = client.detached_list("my-sandbox")

    # Kill a running command
    client.detached_kill("my-sandbox", cmd.id)

Async version:

async with AsyncAgentKernel() as client:
    cmd = await client.exec_detached("my-sandbox", ["python3", "train.py"])
    logs = await client.detached_logs("my-sandbox", cmd.id)

Streaming

for event in client.run_stream(["python3", "script.py"]):
    if event.type == "output":
        print(event.data["data"], end="")

Configuration

client = AgentKernel(
    base_url="http://localhost:18888",  # default
    api_key="sk-...",                  # optional
    timeout=30.0,                      # default
)

Or use environment variables:

export AGENTKERNEL_BASE_URL=http://localhost:18888
export AGENTKERNEL_API_KEY=sk-...

License

MIT

Metadata

Release files for agentkernel-sdk 0.21.1

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for agentkernel-sdk 0.21.1
File Size Uploaded Provenance Actions
agentkernel_sdk-0.21.1.tar.gz 71.8 kB With provenance Details for agentkernel_sdk-0.21.1.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for agentkernel-sdk 0.21.1
File Interpreter ABI Platform Actions
agentkernel_sdk-0.21.1-py3-none-any.whl Python 3 none any Details for agentkernel_sdk-0.21.1-py3-none-any.whl

Total release size: 96.0 kB

Release files / agentkernel_sdk-0.21.1.tar.gz

Download URL agentkernel_sdk-0.21.1.tar.gz
Size 71.8 kB
Tags Source
SHA-256 checksum
How to use checksums
cab415f6d9000ba88d27eaeb31b7add0924fe974cb07221c795830c73897a68d
BLAKE2b-256 checksum
How to use checksums
1304679876984a9ae1c5831ac814a8d542fe1a33c6ff9ede896b81ca686af3a1
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 8, 2026.

Transparency log

Release files / agentkernel_sdk-0.21.1-py3-none-any.whl

Download URL agentkernel_sdk-0.21.1-py3-none-any.whl
Size 24.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
f517f2832e2738345374dbb5b1c8b313af443a0cc7cc4b640c72af87be752bf8
BLAKE2b-256 checksum
How to use checksums
dda876409b45a0609ade7d0cae4d9da6c56fab51a23501aa04bb15eb8edc70fa
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 8, 2026.

Transparency log

Security for agentkernel-sdk 0.21.1

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for agentkernel_sdk-0.21.1.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 8, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.21.1 This release

2 release files Full provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page