OpenClaw AI Agent Builder – Autonomous Agentic Systems (Mac Mini, From Scratch)

Summary
Build and run autonomous OpenClaw agents on our Mac Mini systems — persistent agents that reason, use tools (browser automation, shell, file and email operations), and run on a heartbeat without supervision. You'll build these from scratch and harden them so they can be trusted with real access. All work is done over remote access into our Mac Mini cluster; agents run sandboxed.
What you'll build:

OpenClaw agents from scratch on Mac Mini — provider/model config (openclaw.json), channel wiring, fallback chain
Tool and skill integration (browser automation, shell, file/email ops as scoped)
Heartbeat scheduling and persistent memory
A hardened, sandboxed deployment — non-negotiable (see application requirements)
A README documenting the agent's setup, tools, and permission boundaries

Required: Demonstrated experience building OpenClaw agents specifically — not generic LangChain/AutoGPT agents.
To apply, include all of the following in your first message — anything missing will be ignored:

A video walkthrough (Loom or screen recording) of recent OpenClaw agents you've built from scratch, showing the agent actually running — not slides
In that same video, explain exactly how you keep the agent secure: sandboxing, credential handling, tool and permission scoping, and what the agent is and is not allowed to touch
Which models/providers you run OpenClaw with, and why
Your availability and how many agents per week you can build

Rate: $75–$100 per agent, depending on complexity
Timeline: 3 days from assignment