Thierry Dan

OpenClaw ClawDBot Moltbot WhatsApp Telegram n8n Discord Workflow Setup

TicketLabel • Freelance

Feb 2026 - Mar 2026 • 1 mo

Built a multi-channel automation system for a ticket brokering business using OpenClaw, ClawDBot, and Moltbot. Built Telegram and WhatsApp reply bots for fulfillment dispatch and inventory pricing queries operator sends a message from their phone, agent processes and responds with live data. Connected n8n workflows for event analytics, secondary market insights, and bookkeeping automation. Integrated Discord channel as a secondary alert and command interface. Deployed on a physical Mac Mini with Ollama for local LLM inference, ensuring zero data leaves the machine. Configured systemd for 24/7 uptime with automatic restart on failure. Full documentation delivered for non-technical operator handover.

Python Playwright SQLite LLM Automation Customer.io Batch Campaign

Reach Marketing • Freelance

Feb 2026 - Mar 2026 • 1 mo

Built a stateful autonomous agent for a high-volume educational marketing organization that ran daily segmented campaigns targeting university students across dozens of schools. The client was managing every university in a separate Excel file. That was the first problem I fixed. Migrated all fragmented Excel spreadsheets into a unified SQLite database with a clean relational schema. Built querying logic to filter students by university, major, and a recency window — for example, "not contacted in the last 14 days." The agent pulled the right list every single day without any human input. Built the global context LLM copywriting engine next. Before writing any campaign copy, the agent evaluated the full state of all active campaigns across all schools. It passed this global context to Claude API, which then wrote highly persuasive, major-specific marketing copy for each audience segment. Biology majors at one school got different copy than Kinesiology majors at another. Everything was dynamic and personalized at scale. Built the Playwright headless browser automation layer to log into the client's custom HTML admin backend, navigate the UI, find the right certificate programs, and inject the LLM-written copy alongside updated deadline fields. No human touched the backend interface during a campaign run. Built the Customer.io orchestration layer last. The agent assigned a sequential batch ID to each day's student list, generated the CSV payload with first name, last name, major, and batch ID, logged into Customer.io via headless browser automation, uploaded the CSV, mapped the attributes, and created a broadcast draft segmented to that exact batch. Added fallback logic for schools missing major data those audiences were automatically routed to generalized program selections.

OpenClaw NemoClaw Mission Control Ollama VPS Multi-Agent Deploy

vantadev • Freelance

Feb 2026 - Mar 2026 • 1 mo

Deployed a full production multi-agent system for a technology client who needed OpenClaw, NemoClaw, Mission Control dashboard, and Ollama all running together on a VPS with real security controls and real agents doing real work. Started with the infrastructure layer. Spun up the VPS, installed OpenClaw and connected it to NemoClaw for kernel-level sandboxing and policy-based guardrails. Configured the Ollama local inference layer so all sensitive data stayed on the machine and never hit an external API. Set up the Mission Control dashboard so the client could see all agents, their status, active tasks, and logs from one screen. Built and configured four production-ready agents, each one scoped to a specific use case with its own skill set, tool permissions, and memory profile. Each agent had tools.profile:full enabled real file access, real API calls, real command execution. Not just chat. Connected the agents to external tools and APIs where needed. Integrations included REST APIs, webhook receivers, and a Discord channel for alert routing and command dispatch. Every integration was tested end to end before delivery. Hardened the full stack with NemoClaw sandboxing controls, permission scoping per agent, no exposed ports, and a privacy router that blocked sensitive payloads from leaving the local network. Configured systemd daemons for all core services so the system survived reboots, crashes, and OS updates without manual intervention. Ran a full Zoom walkthrough with the client at delivery. Covered how to add new agents, update skills, read Mission Control logs, and restart services. Wrote a complete handover document so their team could manage the system independently after delivery.