AgentSDK Openclaw Docker

English | 中文

Quick Reference

AgentSDK Openclaw Docker

AgentSDK Openclaw is a highly usable multi-domain Agent framework and service built on OpenClaw, integrating vertical domain Agent capabilities such as code generation, web search, research analysis, and mathematical calculation. This document introduces the Docker image architecture design and usage of AgentSDK Openclaw.

Supported Tags and Dockerfile Usage

Tag Specification

Tags follow the format:

<Openclaw Version>
Field Example Value Description
Openclaw Version 2026.5.22 Corresponding to the official OpenClaw release version identifier

Image Repository Address

AgentSDK Openclaw images are hosted on the Atlas Community Image Repository:

https://www.hiascend.com/developer/ascendhub

Complete Image Examples:

swr.cn-south-1.myhuaweicloud.com/ascendhub/openclaw:2026.5.22

Image Architecture

AgentSDK Openclaw adopts a three-layer image architecture design:

Layer Image Tag Description Dockerfile
Layer 1 openclaw:base-{version} Infrastructure including SSH/uv/bun/pnpm/npm global packages/pip packages/Playwright/Chromium/gosu/ffmpeg/rsync Dockerfile.openclaw-base
Layer 2 openclaw:app-{version} Official OpenClaw build artifacts (multi-stage build) Dockerfile.openclaw-app
Layer 3 openclaw:{version} Custom layer with subagent-coordinator + Hermes + skills Dockerfile.openclaw-overlay

Build Parameters

Parameter Description Required Example Value
VERSION OpenClaw version number No 2026.5.22
REGISTRY Image registry prefix No localhost
OFFLINE Offline mode, use local sources only (default false) No true
SKIP_BASE Whether to skip base image build (default false) No true
SKIP_APP Whether to skip app image build (default false) No true
SKIP_OVERLAY Whether to skip overlay image build (default false) No true
SKIP_PLUGINS Whether to skip plugin preparation (default false) No true
OPENCLAW_SRC OpenClaw source code directory No /path/to/openclaw-src
HERMES_SRC Hermes Agent source code directory No /path/to/hermes-src
DOCKER_REGISTRY_NPM NPM mirror URL No https://registry.npmmirror.com

Quick Start

Build AgentSDK Openclaw Images

Use the unified build script to build images:

# Build all three layers of images with one click
cd AgentSDK/openclaw/docker
bash ./build-openclaw.sh

# Only build the final image (skip base and app, use existing images)
bash ./build-openclaw.sh --skip-base --skip-app

Run AgentSDK Openclaw Containers

Use the deployment script to run containers:

# Quick deployment (single instance)
API_KEY=api-key bash ./scripts/deploy.sh quick -n 1 -m mode-name -u http://xxxx.xx.xx.xx:xxxx -p xxxx -i openclaw:2026.5.22 --skills --name openclaw-test

How to Develop Secondarily

# Use AgentSDK Openclaw image as base image and add user software
FROM openclaw:2026.5.22

# Install additional dependencies
RUN apt update -y && \
    apt install -y --no-install-recommends \
        extra-package && \
    apt clean && rm -rf /var/lib/apt/lists/*

# Install Python dependencies
RUN pip3 install --break-system-packages \
    -i https://mirrors.aliyun.com/pypi/simple \
    extra-python-package

# Copy user-defined skills (deploy script extracts from skills-shared and merge-mounts)
COPY --chown=root:root ./custom-skills/ /home/node/.openclaw/skills-shared/
RUN chown -R node:node /home/node/.openclaw/skills-shared/ && \
    chmod -R 755 /home/node/.openclaw/skills-shared/

WORKDIR /app
CMD ["node", "openclaw.mjs", "gateway"]

Skills Mounting Mechanism

The deploy script automatically merges image built-in skills with host custom skills and bind-mounts the result when starting containers:

Stage Path Description
In image /home/node/.openclaw/skills-shared/ Skills packaged in the overlay layer (extracted at deploy time, shadowed by bind mount at runtime)
Host openclaw-configs/skills-merged/ Merge directory (image skills + host skills union, host takes priority)
Container /home/node/.openclaw/skills/ bind mount from skills-merged/, where Gateway loads skills

Adding skills dynamically (after container is running, no image rebuild needed):

# Option 1: Operate directly on the host merge directory
cp -r ./my-skill openclaw-configs/skills-merged/my-skill/
chmod -R 755 openclaw-configs/skills-merged/my-skill/

# Option 2: Operate via container path (equivalent, since it's a bind mount)
docker cp ./my-skill openclaw-1:/home/node/.openclaw/skills/my-skill/

Restart Gateway after adding skills: enter /restart in the WebUI, or run docker exec openclaw-1 pkill -u node -f gateway (the health monitor will auto-restart it).

Supported Hardware Architectures

Architecture Description
x86_64 Intel/AMD 64-bit architecture
aarch64 ARM 64-bit architecture

License

As with all container images, pre-installed software packages (Python, Node.js, system libraries, etc.) may be subject to their own license restrictions.