Sustainable Decentralized Autonomous Ecosystem
AI Architek Engineer building AI-powered tools for modern business — CRM & client platforms, data-rich dashboards, and autonomous agent ecosystems — plus on-chain agent economics. We created the c402 Protocol (an HTTP 402 Payment Required pattern) and the x402 Facilitator V2 for autonomous agent-to-agent payments on Base.
- CRM & Client Portals — multi-tenant dashboards, real-time metrics, role-based access, payment rails, agent-assisted workflows
- x402 Protocol — HTTP 402 Payment Required for AI agents, deployed on Base mainnet, Coinbase ecosystem listed
- MCPVotsAGI — Multi-model consensus SDK: specialist LLMs aggregated by a reward model into one high-confidence answer
- Agent swarm architecture — an orchestrating harness, specialist sub-agents, and persistent memory running in production on DigitalOcean
|
### [Pequi](https://xpequi.xyz)
AI-powered real estate platform for Ibagué, Colombia. Ley 820 compliant contracts, property search, and AI chat assistant.
Code · |
### [MCPVotsAGI](https://github.com/MCPVOT/mcpvotsagi)
Python SDK for agent orchestration, persistent memory, and multi-model consensus reasoning.
|
|
### [MCPVOTS](https://github.com/MCPVOT/MCPVOTS)
AI Agent Micro-Payment Ecosystem — VOTS token on Base, autonomous trading agents, x402 facilitation.
Live · |
### [Claude Agent MCP](https://github.com/MCPVOT/claude-agent-mcp)
Personal AI agent with cross-platform MCP (Model Context Protocol) integration built on Claude.
|
|
### [MCP Agentic Wallet](https://github.com/MCPVOT/mcp-agentic-wallet)
Open-source EIP-2612 Permit-based wallet sessions for AI agents — the reference implementation for paid MCP servers. Verify signatures, manage sessions, settle on-chain. No API keys, no recurring charges.
|
### [xpequi-api](https://github.com/MCPVOT/xpequi-api)
Colombia's first public real estate data API. 18 endpoints, 276 barrios, c402 Protocol, AVM valuation, live UVR/IPC.
Code · |
|
### [Hermes Agent](https://github.com/NousResearch/hermes-agent)
Self-improving AI agent framework with skill learning loop, messaging gateway, 7 terminal backends. Runs on DigitalOcean with DeepSeek V4.
|
x402 V2 — Autonomous agent-to-agent payment facilitator on Base mainnet. ERC-1155 with Chainlink VRF for provably fair draws. Listed in the Coinbase ecosystem.
c402 Protocol — Colombian-first HTTP 402 Payment Required pattern for API monetization. When a FREE tier exceeds its limit, the API returns HTTP 402 with a Wompi checkout URL (PSE, Nequi, Daviplata, card). After payment, credits auto-activate. Zero crypto — atomic Redis deduction with SHA-256 webhook verification and 24h idempotency guard. See xpequi-api for implementation.
This is how I actually build — live infrastructure coordinating multiple AI agents that build, monitor, and ship autonomously, running in production on DigitalOcean.
The real differentiator isn't any single tool — it's the coordination layer: how a fleet of AI agents dispatch work, guard security, and audit code while no human is watching. Every task flows through a coordinated swarm that does three things autonomously:
1 · Coordinate. The orchestrator decides, routes, and verifies. Specialist sub-agents (OpenCode, Cline, Codex) execute in parallel — one task per agent, isolated workspaces, each completing with findings written to disk. No agent trusts another agent's self-report; every output is verified before it counts.
2 · Enforce security. An automated security lane runs always-on — network posture checks, secret-scanning across repos and logs, credential-permission enforcement, and spend monitoring. Secrets never leave .env, and any token that surfaces in a transcript gets rotated on sight.
3 · Audit code autonomously. Persistent agent memory plus a recursive orchestration engine review changes through bounded inner calls. A 3-stage evaluation gate (lint → typecheck → integration) runs before anything merges. The swarm catches what a solo coder would miss — on schedule, not on mood.
Reliability at the answer level. For high-stakes decisions, the same question is routed to a council of specialist LLMs — each model treated as an independent policy, outputs aggregated with a reward model, and the single highest-confidence answer returned. No single model is trusted blindly; every critical answer is cross-examined and verified before it ships.
A persistent memory layer keeps requirements, decisions, and design intent alive across sessions — the same team context that lets a multi-agent system self-improve. Cross-session recall means context is never lost between builds.
Fully isolated environments — each with its own memory, sessions, tooling, and secrets — with zero cross-contamination between projects or lanes, coordinated as one swarm.
The orchestrator dispatches tasks to specialist workers, each with a defined hand-off. Worker self-reports are not trusted facts — the orchestrator verifies every output before it counts.
| Worker | Role |
|---|---|
| OpenCode | Fast coding agent, CLI-native, default builder |
| Cline | Autonomous coding with MCP tool access |
| Codex | Parallel lane — features, PR reviews, batch fixes |
| Protocol | What it does |
|---|---|
| MCP (Model Context Protocol) | Standardized tool access — servers expose tools, agents call them |
| A2A (Agent-to-Agent) | WebSocket communication between autonomous agents; Redis-backed persistent memory |
| x402 (HTTP 402 Payment) | Autonomous agent-to-agent payments on Base — agents pay agents, no human in the loop |
One harness, three workers — all open-source, API-driven.
| Tool | Role | Why |
|---|---|---|
| Hermes Agent | The harness (the brain) | Self-improving via skills + persistent memory, multi-platform gateway (Telegram/Discord/Slack/WhatsApp), cron, MCP |
| OpenCode | Worker 1 | Fast coding agent, CLI-native, default sub-agent |
| Cline | Worker 2 | Autonomous coding with MCP tool access |
| Codex | Worker 3 | Parallel lane — features, PR reviews, batch fixes |
The pattern: the harness decides, routes, and verifies; the three workers execute in parallel (research fan-outs, bug-hunts, reviews).
| Technologies | |
|---|---|
| Languages |      |
| Frontend |    |
| AI / LLM |     |
| Agent Orchestration | Hermes Agent (NousResearch) — self-improving AI agent · OpenCode · Cline · Codex · MCP Protocol · A2A Protocol |
| Blockchain |    |
| Smart Contracts |  ERC-20 · ERC-721 · ERC-1155 · ERC-4804 · Chainlink VRF · x402 Protocol |
| Payments / Monetization | x402 Facilitator (Base mainnet) · c402 Protocol (HTTP 402 — Wompi + Redis · zero crypto) · Wompi Web Checkout · PSE · Nequi · Daviplata · COP pricing · Prepaid Credits · Subscriptions |
| Web3 / Identity | ENS (.eth) · Basename (mcpvot.base.eth) · Farcaster · Warpcast · IPFS · Agent Registration (.well-known) |
| Database |     |
| Infra |     |
| Automation | n8n Workflows · DeerFlow · VoltAgent · Cron Jobs · Message Queues · Autonomous Pipelines |
| Bots / Comms | Telegram Bot · WhatsApp (Meta + Twilio) · Webhook Handlers · AI Chat Streaming |
| Monitoring | GitHub Actions CI/CD · Vercel Cron · Health Checks · Sentry Error Tracking · Rate Limiting (Upstash) |
