Axern documentation

INFRASTRUCTURE FOR AI AGENTS

Run isolated, stateful, observable AI agents and code workloads with one resource and lifecycle model.

Start in ten minutes Explore the architecture

runc + runscOCI + NydusPython · Go · TypeScript

python service · /svc

Publishing and reaching an HTTP Service with the Axern Python SDK

Sandbox as the primitive

Runs, services, persistent workspaces, and agent tasks share one isolation and cleanup contract.

Durable control plane

Placement, leases, health, retries, and storage intent survive process and node restarts.

Agent evidence loop

Axrun makes tasks, execution, verification, usage, and trajectories reproducible rollouts.

One platform, two entry points

Use Axern when an application needs programmable sandbox lifecycle, processes, files, services, storage, or tunnels. Use Axrun when an agent workflow needs immutable tasks, controlled concurrency, verification, trajectories, usage, and portable evidence.

Install the local stack and run a sandbox through the public gateway.
Choose an SDK for Python, Go, or Node.js applications.
Run a managed rollout when execution needs a reproducible agent evidence loop.

Axern is pre-1.0 and under active development. Evaluate authentication, TLS, runtime isolation, image trust, network policy, quotas, and persistent storage before using it for untrusted multi-tenant workloads.