
Teams managing client projects and VPS-hosted applications often jump between an admin dashboard, terminal commands, payment records, deployment tools, and server configuration. That context switching slows routine work and makes operational details harder to review in one place. Gatekeeperd MCP connects an AI assistant to the Gatekeeperd administration API through the Model Context Protocol (MCP). It exposes tools for project and customer management, balances and payments, invoices, deployments, Docker containers and images, Nginx sites, SSL certificates, audit history, and system status. An operator asks a question or requests an operation in natural language; the assistant selects the relevant MCP tool, receives structured results, and presents them in the conversation. Actions that change project or server state can be reviewed and approved by the operator before execution. The project is for developers, small teams, and service operators who manage multiple web projects on VPS infrastructure and want to work from an AI assistant without repeatedly switching between interfaces. In the demo, the assistant lists projects, identifies outstanding balances, requests a project block with an approval step, and retrieves the running container inventory. The creative element is the bridge between a conversational AI client and a real administrative API: the assistant can answer operational questions and invoke defined backend actions, while Gatekeeperd remains responsible for authentication and execution. The modular MCP tools make the integration extensible across administrative domains. The result is a single conversational entry point for routine project administration and VPS operations, with human review available for consequential changes.
27 Sep 2026

Mikesplore Voice Representative is a realtime voice interface for Michael Odhiambo’s developer portfolio. Visitors can speak naturally and ask about projects, technical skills, articles, achievements, experience, and public resources. The application uses AssemblyAI’s Voice Agent API as the core voice system. AssemblyAI handles realtime speech recognition, turn detection, interruption handling, agent reasoning, tool calling, and spoken audio generation through a single WebSocket connection. The agent uses verified portfolio tools backed by a FastAPI and PostgreSQL backend, ensuring that answers are grounded in real portfolio data instead of invented information. The interface is available as a standalone voice page designed for desktop and Android browsers. Visitors can receive spoken answers and interact with verified project, article, CV, certificate, and media resources through explicit browser cards and links. The project focuses on making a personal portfolio conversational, accessible, and useful without requiring visitors to search through multiple pages manually. Its backend-first design also provides a secure foundation for future authenticated portfolio-management actions while keeping public visitor interactions read-only.
30 Sep 2026

Overview VelaAI is a personal conversational AI agent for Linux desktops that transforms how we interact with our computers. Instead of relying on traditional remote desktop streaming (which suffers from severe bandwidth overhead) or complex SSH syntax (which demands precise technical command knowledge), VelaAI adopts a purely intent-driven approach. You talk to your computer naturally; whether from your phone or a remote network, and it listens and acts. How it Works The platform features a clean, highly secure architectural separation. The lightweight Android mobile application captures raw natural language text or quick-action chips and routes them securely via a local Express proxy. The backend, built with Python 3.13 and FastAPI, runs as a user-level systemd service on the host machine. It passes user prompts to DashScope's Qwen model via tool-calling. Qwen dynamically extracts the user's core intent and maps it to specific, structured parameters. These parameters are then safely executed by modular OS routers interacting with core Linux CLI utilities like playerctl, nmcli, and xrandr Safety First Because system automation demands rigid guardrails, safety is strictly enforced at the backend level. VelaAI is safe by default, using JWT-protected authentication, strict directory whitelists to prevent path traversal, and hard barriers against executing arbitrary shell scripts. Any destructive action, such as killing a process or shutting down the machine, intercepts the execution flow and requires explicit confirmation in the chat thread. It represents a fast, accessible, privacy-first companion built for hands-free workflow automation.
13 Jul 2026