I already have OpenClaw installed on a Mac mini M‑series with 64GB RAM and local models running. Right now, OpenClaw is basically behaving like a single chatbot. I need an expert to: Wire it so it can plan, break down tasks, and call tools/skills autonomously (not just answer one prompt at a time). Set up a small number of core agents and workflows that I can extend later (voice agent, CRM tasks, etc.). Keep things simple enough that once this “brain” is working, I can let OpenClaw handle most of the future build‑out itself. No big infrastructure project, just: take what’s installed and flip it from “chatbot” to “autonomous operator.” My Current State Mac mini M‑series, 64GB RAM / 1TB, dedicated user account for OpenClaw. OpenClaw is installed and runs, dashboard comes up. Local LLMs are available via Ollama (e.g., Qwen / GLM‑4.7‑Flash or similar). It answers prompts, but: Doesn’t really plan multi‑step tasks. Doesn’t call tools/skills reliably. Feels like a fancy chat window, not a “mission control” brain. What I Want You To Do (Tight Scope) Review and Fix OpenClaw Brain Setup Check my current model configuration (local + any cloud I have wired). Ensure OpenClaw is actually using the local models by default and can fall back to cloud only when needed. Turn on / configure whatever is needed so it can: Create plans (multi‑step task graphs). Call internal tools / skills (file access, web, shell, etc.) instead of just chatting. Create 2–3 Concrete “Self‑Running” Use Cases Examples (we can agree together, but think along these lines): A research + write pipeline: given a topic, it plans, searches, drafts, and cleans a document without me micromanaging each step. A CRM / business ops helper: can read a brief, then propose and execute steps like drafting emails, messages, or internal SOPs. A lightweight agent manager: given a goal, it decides which internal skill/agent to use and runs it end‑to‑end. The key: I give it a goal, not a line‑by‑line set of instructions, and it executes. Prep for Future Voice & Phone Agent (No Build‑Out Yet) Make sure the architecture is ready so I can later attach: A voice / phone agent. External tools (CRMs, calendars, etc.). You do NOT need to build the telephony stack now; just: Confirm the right skills/tools structure. Show me where future “voice agent” logic will plug in. Documentation & Mini‑Handoff A short, clear README that covers: How to start/stop OpenClaw and any background services. How to define a new skill/tool. How to define a new agent that can plan and act, not just chat. Optional short screen recording walking through: “Here’s a goal we give it.” “Here’s how it plans, calls tools, and finishes without more prompts.” Ideal Freelancer Hands‑on experience with OpenClaw or similar agent frameworks (LangGraph, crewAI, etc.). Strong understanding of tool‑calling / planning / multi‑step agent flows. Comfortable in macOS terminal, Node/TypeScript or Python, and configuring env files. Bonus: experience with local LLMs on Apple Silicon (Ollama, GLM‑4.7‑Flash, Qwen, etc.), but the priority is getting the agent logic working. Budget / Expectations This is a tight‑scope configuration and wiring job, not a huge greenfield project. I expect: 1 working autonomous “brain” configuration. 2–3 demo goals that it can handle end‑to‑end. Enough docs so I can extend it and later let OpenClaw build voice agents and other workflows on top.