Inference APIs / OpenAGI / Lux alternative

OpenAGI Lux alternative

OpenAGI Lux is a computer-use model and SDK stack. It observes screenshots and produces actions in a repeated observation/action loop. Model inference, the desktop being controlled and the persistent machine hosting your application are separate parts of the system.

Start buildingCompare capabilities ↓

Reviewed 2026-10-18 · Published by the Sandbox0 team

When OpenAGI / Lux fits

Lux fits developers who want computer-use model inference with SDK control loops and desktop-action adapters. Its Actor and Thinker models offer different execution styles while using the same published token-price table.

When to consider Sandbox0

Consider Sandbox0 when the agent application needs installed Linux tools, retained configuration and files, and a whole RootFS that expands online. Keep a model service such as Lux if appropriate, and test the required display, input and screenshot tooling rather than assuming raw sandbox compute replaces the model.

Capabilities

OpenAGI / Lux and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeOpenAGI / LuxNot established in this reviewSandbox0ttl and hard_ttl default to 0 (disabled). Configure idle pause and deletion policies for the same persistent sandbox identity; an idle running sandbox still accrues compute charges.Read documentation →
Persistent filesOpenAGI / LuxThe SDK/model service does not establish a hosted durable disk lifecycle.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSOpenAGI / LuxComputer-use inference does not establish retention of an entire writable Linux RootFS.View evidence ↗Sandbox0Persistence extends beyond /workspace to paths such as /usr/local, /etc, and /home. Runtime tmpfs and virtual filesystems are excluded.Read documentation →
Release compute, keep stateOpenAGI / LuxA screenshot/action loop is not arbitrary-machine pause/checkpoint behavior.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateOpenAGI / LuxArbitrary Linux process-memory capture and restore are not established.View evidence ↗Sandbox0Experimental memory checkpoints can preserve compatible process state. Use memory: true on both pause and resume on a compatible deployment. A filesystem checkpoint alone does not preserve running processes.Read documentation →

Online RootFS growth

On growth-enabled images, Sandbox0 automatically expands the writable RootFS as free space runs low, within a configured sparse logical ceiling up to 64 TiB. Agent processes keep running. Growth is enabled in the qualified Cloud region; confirm activation for other regions. Existing images and snapshots retain their captured storage geometry. The ceiling is logical capacity, not preallocated host disk.

Lux inference and screenshot/action adapters do not establish an automatically expanding Linux RootFS. The execution host determines whether installed packages, configuration and accumulated outputs survive. Sandbox0 supplies that separate durable runtime layer.

Files while compute is paused

Sandbox0's Files API reads persistent files without resuming compute. Filesystem-only pauses also permit edits; retained-memory pauses are read-only and reject mutations with 409. SSH, SFTP, file watching, and runtime-only mounts require running compute.

The reviewed Lux docs do not establish a paused Linux Files API. Retain necessary artifacts and task progress on the execution/storage host; model requests do not themselves make a paused desktop’s filesystem editable.

See paused file operations →

Actor and Thinker change the loop, not the storage contract

The model guide presents lux-actor-1 for responsive, direct action loops and lux-thinker-1 for more involved planning. Its examples pass screenshot and action-handler objects to the SDK. Those model choices do not establish persistent disk, a long-lived hosted desktop or process-memory restore.

Source 1 ↗

Desktop integration is part of the deployment

The quickstart documents host-specific input and screenshot requirements, including Wayland setup and operating-system permissions. A browser or desktop must be available for the loop to observe and control. Validate those dependencies in the actual execution image and keep important progress in durable application state.

Source 1 ↗Source 2 ↗

Input and output tokens are separate meters

The developer pricing page lists $0.10 per million input tokens and $0.90 per million output tokens for all Lux models. Repeated observations and generated actions can change token usage per task. These are inference rates, not a fixed price for an hour-long desktop, a completed workflow or retained Linux storage.

Source 1 ↗

Pricing

Compare billing models, then price your workload

CPU allocation, memory, billing minimums, disk, plan fees, and supported workload differ. These meters are not equal-performance machine quotes. Confirm region, resource units, and included allowances before estimating a total.

OpenAGI / Lux

Pricing reviewed 2026-10-18

Lux model input
$0.10 per 1 million input tokens
The published table states the same pricing for all Lux models.
Rate source ↗
Lux model output
$0.90 per 1 million output tokens
Output and input have separate rates; task cost depends on actual usage.
Rate source ↗
Published new-user inference credit
$10 once on registration
A developer-documentation offer for inference, not free desktop runtime or a recurring monthly allowance.
Rate source ↗

Token pricing does not include a documented equal-resource Linux compute or durable-disk allocation. Browser/desktop hosting, model usage and the surrounding agent runtime need separate budgets. The SDK’s task descriptions and marketing benchmark examples are not guaranteed prices per completed task.

Check current OpenAGI / Lux pricing ↗

Sandbox0

$0.015/GiB-hour compute (billed by the second at approximately $0.000004167/GiB-second), plus $0.02/GiB-month of retained incremental RootFS and memory checkpoint data. Shared base images and platform recovery retention are excluded; paused sandboxes accrue storage charges only.

Compute / GiB-hour
$0.015
Retained data / GiB-month
$0.02
Paused compute
$0.00

A 4 GiB agent, 100 running hours

$6.00 compute

4 GiB × 100 hours, billed by the second. Keeping 20 GiB of billable incremental data for a full billing month adds $0.40. CPU is derived from memory by the platform; choose a tier that fits your workload.

Understand storage billing →

Before migrating from OpenAGI / Lux

  1. Inventory Lux models, SDK/action/screenshot adapters, display permissions, state files and actual input/output token use.
  2. Move compatible agent code and Linux dependencies to a Sandbox0 template, retaining the inference endpoint and validating the desktop/browser tooling.
  3. Test control-loop recovery, retained system packages, RootFS growth, paused file access and complete inference plus runtime/storage costs.

Common questions

Is Sandbox0 a drop-in OpenAGI / Lux replacement?

Evaluate API compatibility and product scope. These products serve different core workflows, so a change may require a different architecture.

Does keeping files also keep a running agent?

Filesystem state and volatile process state are separate. Sandbox0 requires explicit compatible memory checkpoints to preserve running processes; its default filesystem resume starts fresh processes.

Sources and review scope

The Boat team's Battleships directory is the discovery source. References below describe the product and plan discussed on this page. Check current availability, regional terms and billing units before migrating. Where the cited material does not establish a capability or price, the comparison identifies that uncertainty.

Battleships research record ↗developer.agiopen.org/docs/models ↗developer.agiopen.org/docs/quickstart ↗developer.agiopen.org/ ↗developer.agiopen.org/docs/pricing ↗Sandbox0 RootFS capacity →

Related comparisons

Baseten Model Serving / Deployments vs Sandbox0 →fal Model APIs, Serverless and Compute vs Sandbox0 →Pipeshift Inference Platform vs Sandbox0 →PoplarML Model Deployment vs Sandbox0 →Replicate Model APIs vs Sandbox0 →RunAnywhere Wally and Local SDKs vs Sandbox0 →