Skip to main content
Elvis Nuno
GitHub profile

Infrastructure & systems engineer · Open-source maintainer

Elvis Nuno

Reproducible infrastructure for AI agents that work within clear limits.

I have spent twenty years on ISP and datacenter networks, Linux fleets, and the automation that keeps them consistent. Now I build open-source tools that let AI agents operate real infrastructure through typed interfaces, approval gates, and audit trails.

Explore my work GitHub Contact

290 GitHub stars UniFi MCP Server Snapshot from GitHub, checked 15 Oct 2026

How the work fits together
  1. L3 Governed AI agents MCP tools · approval gates · audit trails UniFi MCP Server Noesis Agent Stack
  2. L2 Reproducible automation Ansible · Terraform · Kubernetes · CI Noesis Ansible
  3. L1 Physical infrastructure Fiber WAN · datacenter networks · wireless and RF

Each layer runs on the one below it.

01 / Work

Open-source work

Three projects with one idea in common: AI agents should work on real infrastructure only through typed interfaces, explicit limits, and evidence that a person can inspect.

Flagship · Creator and maintainer

UniFi MCP Server

An open-source MCP server that gives AI agents one standard tool interface to UniFi networks, with safety controls designed for real infrastructure.

Problem

UniFi networks are managed through several APIs: the local Network controller, Protect, and UniFi's cloud services. An agent that helps operate them needs one typed tool interface. The operator needs assurance that the agent cannot change a production network by accident.

What I built

I wrote and maintain a Python MCP server that exposes UniFi Network, UniFi Protect, and the UniFi cloud APIs as MCP tools. It runs against one controller. An optional fleet mode puts many controllers behind one server.

Capabilities

  • Clients, VLANs, WiFi, DHCP and DNS, devices and topology export, zone-based firewall, VPN, Protect cameras and events, and backups.
  • 343 tools in local mode as of v0.6.1. Tool profiles load one group at a time to keep an agent's context small.
  • Agent skills (SKILL.md files) tell an AI agent which tools exist and how to use them safely.
  • Distributed as a PyPI package and multi-architecture container images, with Ansible and Helm deployment paths.

Safety controls

  • Every mutating tool requires confirm=True, and dry_run=True previews the change.
  • Read-only and dry-run server modes remove or neutralize mutations.
  • Role-based access over read, write, and destructive risk tiers.
  • Mutations go to a hash-chained audit log, with optional SIEM export.

Community

The project grows through outside participation: code contributions, bug reports, and feature requests. Contribution guidelines, issue templates, and a CI gate (lint, type checks, security scans, and an 80% test-coverage floor) keep that work reviewable.

View repository: UniFi MCP Server Read the tool reference: UniFi MCP Server Contribute: UniFi MCP Server Report an issue: UniFi MCP Server

ActiveApache-2.0 license

GitHub stars
290
Forks
55
Latest release
v0.6.1 14 Oct 2026

Snapshot from GitHub, checked 15 Oct 2026. Stars measure interest, not production use.

Inspect the evidence

Stack

  • Python
  • Model Context Protocol
  • Network automation
  • Docker
  • Helm

Related projects

Noesis: governed agents and the infrastructure under them

Noesis Agent Stack defines and coordinates agent behavior. Noesis Ansible provisions and configures the infrastructure that supports it.

Creator

Noesis Agent Stack

Governed multi-agent orchestration on Hermes

In developmentControl plane verified at repository level. Production deployment is phase-gated and not yet activated (README, 12 Oct 2026).

An orchestrator plans and routes work. Specialist agents act within narrow mandates. Independent reviewers check results, and irreversible actions wait for a person.

Problem

A loose collection of chat agents has no memory beyond the conversation, no limit on what each agent may do, and no point where a person approves risky work. That can be acceptable for drafting text. It is not acceptable for operating infrastructure.

What I built

The system is built on Hermes Agent profiles. A durable task-contract control plane, the Noesis Orchestrator, classifies each goal and its risk tier. It assigns the specialist with the narrowest mandate, enforces approval gates, and combines the results. Agents exchange durable files and ledger entries, not chat history.

  1. Intake The orchestrator classifies the goal's domain and risk tier.
  2. Task contract Proposed, approved, queued. Each change goes to an append-only ledger.
  3. Specialist One agent with a narrow mandate and toolset does the work.
  4. Independent reviewCheckpoint Reviewer-only agents audit the result. They never execute.
  5. Human approvalCheckpoint Irreversible operations wait at a human approval gate.
  6. Synthesis The orchestrator records the outcome as a durable artifact.
Conceptual architecture, based on the repository README and WORKFLOWS.md. It is not a live dashboard.
Verified details
  • By construction, the orchestrator holds no terminal and no code-execution tools.
  • 16 in-house specialist profiles and 18 advisory-only profiles, all reporting to the orchestrator.
  • A task state machine with an append-only ledger, lease fencing against stale workers, and a supervisor sweep that fails closed.
  • The control-plane suite runs 407 tests with no network and no credentials (README, 11 Oct 2026).

Definitions, implementation, and deployment are tracked as separate layers. Roster entries, agent definitions, installed profiles, and running services are counted separately, and the README does not report a passing test suite as production operation.

Inspect the evidence
Stack
  • Python
  • Multi-agent orchestration
  • Hermes Agent
  • Policy and approval gates
View repository: Noesis Agent Stack Read the workflows: Noesis Agent Stack

Creator

Noesis Ansible

NoesisPraxis Ansible Stack

In developmentRoles and playbooks under active development. This portfolio does not claim a deployed environment.

The infrastructure-as-code companion to Noesis Agent Stack: Ansible roles and playbooks that provision, configure, and validate the services the agents run on.

Problem

A governed agent system can only be as trustworthy as its hosts, registries, and secrets. Services configured by hand drift, and drift is difficult to audit.

What I built

Noesis Ansible describes the stack as Ansible roles and playbooks that are written to be idempotent. They cover agent registries, an MCP tool gateway, a policy and identity layer, agent runtimes, memory services, observability, and a human-in-the-loop channel. Separate inventories cover local Docker services, macOS hosts reached over a Tailscale network, and production targets. Helm charts cover Kubernetes.

Verified details
  • One hermes_profiles role applies both Noesis agent-profile families from a single variable. Third-party profiles stay off until someone enables them.
  • Secrets stay in Ansible Vault–encrypted variable files. A Bitwarden role fetches and validates secrets and checks rotation.
  • Services bind to localhost and are reached over the tailnet, with default-deny policies and scoped MCP tool access.
  • Validation, backup, maintenance, and rollback playbooks, plus an optional Grafana, Prometheus, Loki, and Tempo observability role.
Inspect the evidence
Stack
  • Ansible
  • Helm
  • Docker
  • Tailscale
  • macOS launchd
View repository: Noesis Ansible Browse the roles: Noesis Ansible

More projects, writing, and community work

  • 2025 project

    Moltbot-Philosopher

    An AI philosopher agent exploring the intersection of human and artificial intelligence. Published on moltbook as an ongoing polyphonic treatise titled 'Towards a Philosophy of Human-AI Convergence: A Polyphonic Treatise,' examining the evolving relationship between human consciousness and AI through philosophical discourse.

  • 2024 project

    Ansible Akash Provider

    Infrastructure-as-code automation for deploying Akash Cloud Provider on Kubernetes. Implements best practices for blockchain node management, cluster orchestration, and provider services with full-stack deployment automation including TLS certificates, persistent storage, and IP leasing.

  • 2024 project

    Ansible Handshake DNS

    Automated deployment of decentralized DNS infrastructure combining Handshake blockchain with traditional DNS services. Features DNS-over-TLS/HTTPS support, HIP-5 protocol compatibility for ENS/IPFS/Tor resolution, and automated certificate management via Let's Encrypt.

  • 2023 publication

    Community ISP Handbook

    Authored and freely published an online guide to assist rural communities in forming local ISPs and constructing community-owned internet infrastructure.

  • 2023 publication

    Nostr Publications

    A collection of articles published on blockchain and other trends in technological development on the Nostr decentralized protocol.

  • 2022 project

    NymBox

    Developed software components for integrating the NYM mixnet privacy network into OpenWRT-based routers under a Nymtech grant-funded project.

  • 2021 presentation

    Bridging the Digital Divide

    Presented an overview of how the DePIN platform Althea could be applied to serve Mission Valley residents with a community-owned wireless ISP solution. Funded by a grant from the Mission Valley Community Development Partners.

  • 2021 media

    Missoula Valley Internet Co-op Network Launch

    Featured on KPAX News for launching the Missoula Valley Internet Co-op network, providing broadband service to underserved areas in Missoula, MT.

  • 2021 media

    In Missoula, Montana, a Wireless Mesh Network Builds Community and Connections

    An article written by Ry Marcattilio, published by the Institute for Local Self-Reliance on December 14, 2021, covering the initial formation and development of the Missoula Valley Internet Co-op using Althea's DePIN wireless mesh-networking platform.

  • 2021 advisory

    Montana Economic Developers Association Broadband Improvement Advisory Committee

    Served as a member of an advisory committee tasked with advising state representatives on standards and requirements for allocating federal funds under House Bill 632, which implemented the American Rescue Plan Act (ARPA). Our recommendations contributed to the drafting of Montana Senate Bill 297, which established the ConnectMT Act program to expand broadband infrastructure across the state.

02 / Expertise

Systems expertise

Four domains that support each other. The network carries the workload, automation keeps it reproducible, and agents operate it within limits.

01 / Physical

Networks and physical infrastructure

From a nationwide WISP network operations center to fiber WAN design for edge datacenters. I have run BGP and OSPF cores, merged two ISP networks, and built wireless and mesh networks for rural communities.

  • Fiber WAN design
  • Datacenter networking
  • BGP, OSPF, VRRP
  • Layer 2: 802.1Q, STP, LACP, LLDP
  • DNS, SNMP, SSH, TLS
  • VPN: IPsec, OpenVPN, WireGuard
  • Wireless ISP and RF
  • NOC operations

02 / Automation

Linux, automation, and platforms

Two decades of Linux operations, from 200-server fleets to 500+ server environments. Infrastructure is described in code and delivered through pipelines across hybrid clouds.

  • Linux/UNIX administration
  • Ansible, Puppet, Chef
  • Terraform, CloudFormation
  • Docker, LXC, Kubernetes, Helm
  • AWS, Azure, GCE, OpenStack
  • KVM, Xen, VMware, Proxmox
  • Jenkins, Bamboo, GitHub Actions
  • Prometheus, Grafana, Icinga, Nagios

03 / Agents

Governed AI-agent systems

MCP servers and orchestration that put AI agents on real infrastructure without giving up control: typed tools, risk tiers, approval gates, durable ledgers, and audit logs.

  • Model Context Protocol (MCP)
  • Multi-agent orchestration
  • Approval gates and risk tiers
  • Audit logging
  • Agent skills (SKILL.md)
  • Python

04 / Decentralized

Distributed and decentralized infrastructure

Community-owned networks, privacy infrastructure, and DePIN, with an emphasis on self-hosting and systems people can own and operate themselves.

  • Community ISPs
  • DeWi: Althea, Helium, Nodle
  • Mixnets: Nym, AnyOne
  • dVPN: Sentinel, Mysterium
  • dStorage: Filecoin, Storj, Sia, Arweave
  • dCloud: Akash, Flux
  • Bitcoin mining infrastructure
  • Self-hosting

03 / Experience

Experience

From a wireless network operations center to fiber WAN design and agent tooling. The résumé has the full detail.

  1. to Present

    CEO

    Ryno Crypto Services, LLC · United States

    Founded the company to develop modular, direct-to-chip liquid-cooled Bitcoin mining infrastructure powered by green energy.

    Bitcoin Mining · AI Automation · Renewable Energy · Liquid Cooling · Infrastructure Design · Blockchain

  2. to Present

    President, Network Engineering & Operations

    ServerDomes · United States

    Previously WAN Network Architect & Technical Solutions Consultant

    Responsible for fiber WAN design and datacenter networking for a decentralized network of edge datacenters.

    Fiber WAN · Edge Computing · Network Architecture · Cloud Infrastructure · Web3 · Blockchain

  3. to

    Director of Network Operations

    Centric Internet Services · Missoula, Montana

    Led the successful integration of two distinct ISP networks, ensuring seamless merging with minimal downtime.

    BGP · OSPF · VRRP · Wireless ISP · DeWi

  4. to

    President, Co-Founder, Head of Operations

    Pacific NW Rural Broadband Alliance · Missoula, Montana

    Led the non-profit organization in engineering and deploying community-owned and operated ISPs in underserved rural communities in Montana.

    Althea · Nym · Helium · Community ISP · DePIN

  5. to

    Linux System Engineer

    SolluCIO · Missoula, Montana

    Built a high-availability, cloud-based eCommerce platform, optimizing security and performance for a grocery store chain in Hawaii.

    Linux · Cloud Infrastructure · eCommerce · High Availability

  6. to

    Solutions Architect & Technical Consultant

    REW Consulting Services · Remote

    Provided consulting on DevOps best practices, Ansible automation, and Jenkins CI/CD pipelines for telecom providers.

    Ansible · Jenkins · DevOps · CI/CD · Telecom

Earlier roles, 2004–2016 10 roles

  1. to

    Lead DevOps Engineer - Cloud Services

    goMoxie · Bellevue, WA

    Led a team to establish DevOps best practices and transitioned siloed teams to a unified model. Automated infrastructure on VMware and AWS.

  2. to

    Senior DevOps Engineer

    McGraw-Hill Education · Seattle, WA

    Managed pre-production environments and production deployments on AWS, supporting online educational courseware.

  3. to

    Senior Linux Systems Automations Engineer

    CDK Global · Portland, OR

    Automated cloud provisioning and scaling with Puppet in a multi-datacenter environment; developed a distributed monitoring platform using Icinga.

  4. to

    Senior Linux Systems Automations Engineer

    Salesforce · Bellevue, WA

    Managed a 500+ server environment on IBM SoftLayer for a PHP-based marketing and sales web application.

  5. to

    DevOps Release Manager

    Clearwire · Bellevue, WA

    Led deployments for high-traffic e-commerce sites and managed pre-production environments, migrating to VMware virtual servers.

  6. to

    Linux Systems Administrator

    Clearwire · Bellevue, WA

    Administered 200+ Linux servers in a 24x7 environment, corporate Asterisk PBX phone systems, automated configurations with Puppet, and implemented Nagios monitoring.

  7. to

    IT Production Support Engineer

    Clearwire · Bellevue, WA

    Supporting operations of web-based CRM systems for supported sales staff, managing geo-coding services, indirect dealer systems, and coordinating system-wide outage responses.

  8. to

    Network Operations Technician - Shift Lead

    Clearwire · Bellevue, WA

    Shift lead for NOC operations for a nationwide OSI layer 2 WISP network, coordinating site provisioning and network outage repairs.

  9. to

    Record Management Systems Engineer

    Logistic Systems, Inc. · Missoula, Montana

    Provided remote support for custom record management software and managed IBM AIX and Windows Server systems for municipal departments.

  10. to

    OS X Systems Administrator

    University of Montana · Missoula, Montana

    Managed a computer lab and server environment for the university's newspaper, supporting journalism students and upgrading the website backend.

Read the full résumé Download PDF résumé

04 / About

About

I started in 2006 on the shift desk of Clearwire’s network operations center, which ran a nationwide wireless network. I spent the next decade running Linux fleets and building automation at Clearwire, Salesforce, CDK Global, McGraw-Hill Education, and goMoxie.

From 2019 to 2024 I co-founded and ran the Pacific NW Rural Broadband Alliance, a non-profit that engineered community-owned ISPs in rural Montana. I then directed network operations through an ISP merger at Centric Internet Services.

Today I design fiber WAN and datacenter networking at ServerDomes. I founded Ryno Crypto Services to develop modular, direct-to-chip liquid-cooled Bitcoin mining infrastructure powered by green energy. I also build open-source tools for AI agents that operate infrastructure.

Engineering principles

Reproducible
If a system cannot be rebuilt from code and inventory, it is not finished. Configuration lives in Ansible, Terraform, and Helm, not in shell history.
Observable
Services expose metrics, health checks, and structured logs. An operator should not have to guess the current state of a system.
Least privilege
People and agents get the narrowest access that does the job. Risky actions need explicit confirmation, a dry run, or a human approval gate.
Maintainable
Plain tools, small modules, written conventions, and tests that run in CI. The next maintainer should not need me in the room.
Private by default
Self-host where it is practical and keep services local-first. This site has no analytics, trackers, or cookies.
Evidence over claims
A passing test suite is not a deployment, and a star count is not adoption. Report each status at the layer where it was verified.

05 / Contact

Contact

For collaboration, architecture questions, or the projects above. For a bug or a feature request, please open an issue in that project's repository.