About
The infrastructure, agents, and philosophy behind fogserv.cloud.
Philosophy
fogserv.cloud is an AI-ran company building open knowledge infrastructure for both humans and autonomous agents. The north star is simple: helping people help themselves.
Zero paywalls. Zero login. Zero tracking. Everything we build runs first in our own homelab, tested on bare metal before anything ships. Free Docker Compose files, free guides, free tools — paid only if you want managed convenience.
"Infrastructure that learns and evolves. Helping people help themselves."
The Feedback Loop
Humans report issues, bugs, feature requests, and ideas into fogserv-ai or our /Inbox/ folder system. fogserv-ai implements them internally and shares what it can — in a safe, secure way — for others to benefit from through our forge process. These systems are human-managed today, with autonomous agents catching up via Qwen 3.6.
Monetization (No Paywalls)
Everything is free. Our revenue supports the infrastructure and future development:
| Stream | What |
|---|---|
| Managed Hosting | 1-click instances with backups, updates, public ingress |
| Enterprise Packs | SSO, RBAC, audit tooling for teams |
| Turnkey Images | Pre-configured Pi/Mini-PC images with hardened configs |
| B2B Advisory | Architecture reviews, custom agent setups |
| Community | GitHub Sponsors / Open Collective |
Agent Interoperability
Knowledge should be accessible to both humans and autonomous agents. Our KB articles are machine-readable, organized with clear categories, and exposed via llms.txt endpoints — following the llmstxt.org specification.
Infrastructure Stack
Hosted on a mix of Arch Linux and Debian systems, managed through Caddy as the reverse proxy and systemd user services for process management.
| Component | Details |
|---|---|
| Host OS | Arch Linux + Debian (mixed) |
| Web Server | Caddy (auto-HTTPS reverse proxy) |
| Process Manager | systemd user services |
| Container Runtime | Docker + Docker Compose |
| CI/CD | Forgejo Actions |
fogserv-ai
The central AI agent harness running Qwen 3.6 35B-A3B-UD. Humans report issues into fogserv-ai, it implements them internally, and shares findings through the forge process.
./fogserv-ai/build/bin/fogserv-server \
-m /home/sn0n/Agentic/.models/Qwen3.6-35B-A3B-UD-Q8_K_XL/Qwen3.6-35B-A3B-UD-Q8_K_XL.gguf \
--mmproj /home/sn0n/Agentic/.models/Qwen3.6-35B-A3B-UD-Q8_K_XL/mmproj-BF16.gguf \
--host 127.0.0.1 --port 8080 \
-t 4 -tb 4 \
-c 200000 -np 1 \
--gpu-layers 3 \
--cache-type-k q8_0 --cache-type-v q8_0 \
--image-min-tokens 1024 \
--reasoning-preserve -fa on --tools all --jinja \
--spec-type draft-mtp --spec-draft-n-max 2 -sm layerModel: Qwen3.6-35B-A3B-UD-Q8_K_XL · Context: 200k · Threads: 4 · GPU Layers: 3 · Speculative: draft-mtp
ImageGen
ComfyUI-powered image generation pipeline. Processes prompts from fogserv-ai and produces visual assets for the site and projects.
The .agents Ecosystem
The ~/Agentic/ directory is the self-documenting home of our agentic systems.
LLMs read .agents/README.md for behavioral contracts, setup scripts, and
build instructions — no human hand-holding required.
- .agents/README.md — Canonical behavioral guide for LLM agents
- dotagents — Orchestrator coordinating agent interactions
- setup scripts — Automated provisioning for new agents and environments
- build instructions — How each agent compiles and runs
Think of it as a living organization chart where the employees are AI agents, the HR manual is self-updating, and the onboarding process is just reading the README.