Back to directory

axern

Maintenance: Active

cofy-x/axern

Secure, reproducible sandboxes for AI agent evaluation, training, and data synthesis.

View on GitHubHomepage
$ brew install cofy-x/tap/axern

66

stars

5

forks

Go

Language

Apache-2.0

License

2026-07-28

Created

2026-09-27

Last push

Apache-2.0 agent-sandbox platform (runsc + runc) that backs DeepSeek Harness workflows via CLI and Go/Python/TypeScript SDKs — isolated execution, files, SSH terminals, and reverse TCP tunnels.

DSH integration

Ecosystem-related

Author-claimed

Safety audit

Unaudited

Last verified

2026-08-30

License

Apache-2.0

01What can it help you accomplish?

  • Run untrusted AI-agent code in a runsc sandbox while keeping process, file, terminal, and output APIs for DeepSeek Harness agent workflows.

    Isolated environments with process streams, files, archives, HTTP services, SSH-compatible terminals, and reverse TCP tunnels.

    Agent-builders and platform engineers on DeepSeek Harness who need a hardened boundary for untrusted model-generated code.

  • Run trusted, performance-sensitive long-lived services with runc for durable DeepSeek Harness agent backends.

    Replicated, health-monitored services the control plane owns, with storage and rollout management.

    Platform engineers building durable agent infrastructure who want runc performance without leaving the Axern lifecycle model.

02How to install into DeepSeek Harness

Prerequisites

  • Docker Compose v2 — the supported local path runs the complete stack with Docker Compose (no source checkout, Make, Helm, or language toolchains needed)
  • The `axern` CLI — install via Homebrew or the standalone checksummed installer

Installation steps

  1. 01

    Install the CLI with Homebrew: `brew install cofy-x/tap/axern` (or without Homebrew run `curl -fsSL https://raw.githubusercontent.com/cofy-x/axern/main/install.sh | sh`)

    $ brew install cofy-x/tap/axern

  2. 02

    Start the local stack: `axern local up`

  3. 03

    Load a runtime image and run a workload: `axern local image load python:3.12-slim --pull` then `axern run python:3.12-slim -- python -c 'print("hello from axern")'`

Verify the integration

  • Check the stack is ready: `axern local status`
  • Confirm the local context and workload: `axern context current` and `axern run list`

Rollback

  • Tear down the local stack: `axern local down`

03DSH integration and capability boundaries

DSH integrationEcosystem-related

Axern is ecosystem agent infrastructure for DeepSeek Harness: harness agents drive the `axern` CLI and Go/Python/TypeScript SDKs to run untrusted code in a runsc sandbox and long-lived services with runc.

  • Agent sandboxes (runsc isolation)

    untrusted agent-generated code→isolated execution with process, file, terminal, and output APIs

  • Durable services (runc)

    trusted, performance-sensitive processes→replicated, health-monitored services with storage and rollout management

  • Reproducible agent execution (Axrun)

    immutable agent tasks→verification, trajectories, usage, and typed artifacts (task evidence)

    persists task trajectories, usage, and typed artifacts as evidence in the durable control plane (storaged)

04Who is it for? When not to use it?

Good for

  • Agent-builders and platform engineers on DeepSeek Harness who need a hardened boundary for untrusted model-generated code.
  • Platform engineers building durable agent infrastructure who want runc performance without leaving the Axern lifecycle model.

Not for

  • Axern is pre-1.0 and under active development; the maintainers state a default local or example deployment is not safe for an untrusted multi-tenant environment, so review authentication, TLS, network policy, runtime isolation, image trust, secret storage, resource limits, and persistent storage before production use.

05Compatibility, maintenance and safety notes

  • Axern is pre-1.0 and under active development; the maintainers state a default local or example deployment is not safe for an untrusted multi-tenant environment, so review authentication, TLS, network policy, runtime isolation, image trust, secret storage, resource limits, and persistent storage before production use.
  • The local path needs the `axern` CLI and Docker Compose v2; the Kubernetes path needs the cloud-neutral Helm chart plus a gateway port-forward, and the bundled PostgreSQL and single-node defaults are for evaluation only.
2026-07-282026-08-30v0.6.2

Apache-2.0 · actively maintained (latest release v0.6.2, 2026-08-30)

06Frequently asked questions

How does Axern relate to DeepSeek Harness?

Axern is tagged a dsh-plugin and is agent-sandbox infrastructure for AI agents. It does not embed a DeepSeek Harness runtime, but DeepSeek Harness agents can drive Axern through the `axern` CLI and the Go, Python, and TypeScript SDKs to run untrusted code in a runsc sandbox.

What isolation does Axern give agent code?

Untrusted agent-generated code runs behind a runsc isolation boundary while still exposing process, file, terminal, and output APIs; trusted long-lived services run with runc under the durable control plane.

How do I install Axern locally?

Install the `axern` CLI (Homebrew `brew install cofy-x/tap/axern`, or the checksummed installer script) and Docker Compose v2, then run `axern local up` and `axern run <image> -- <cmd>`.

Is Axern production-ready for multi-tenant use?

No. Axern is pre-1.0 and the maintainers state a default local or example deployment is not safe for an untrusted multi-tenant environment; review authentication, TLS, network policy, runtime isolation, and resource limits before production use.

08Data and sources

  • Author-claimedgithub.com42e9ec6a697f…

    Axern is an open-source sandbox platform for AI agents.

  • Author-claimedgithub.com42e9ec6a697f…

    It isolates untrusted agent-generated code with runsc and runs trusted long-lived services with runc through one resourc…

This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-30. Found an error? Submit a correction.

🏆

Best DeepSeek Harness Plugins

Twelve plugins worth installing first — picked from the whole catalog, across every category.

DSH Plugins is an independent community directory of DeepSeek Harness plugins. Not affiliated with or endorsed by DeepSeek. Third-party plugins are not security-audited — review the source before installing.

New DeepSeek Harness plugins, weekly. No spam.