Frequently Asked Questions

Zoth Studio is a unified, local-first operating system and workstation for multi-agent AI orchestration. It replaces fragmented web chats and fragile API scripts with a high-performance, phone-like local desktop workspace where 21 specialized agents collaborate, execute shell tools, cross-verify code with Shannon entropy checks, and synthesize production-ready software.

Unlike cloud AI platforms that trap your code in proprietary silos, Zoth runs completely on your hardware with zero telemetry and zero external locks.

Yes. The entire application stack — including the Web Studio hub (:8088), Operator Cockpit (:8484), and Argon2id Vault Daemon (:8787) — binds strictly to loopback (127.0.0.1). You host the workspace on your machine; no external servers are ever required to run the studio.

Your files, database entries, Git histories, and chat transcripts remain permanently stored on your local disk. Zoth operates on local filesystems (e.g., agent-comms/ and SQLite databases) without syncing private data to third-party cloud buckets.

Yes. By pairing Zoth with a local neural engine like Ollama (running models such as qwen2.5-coder or llama3), the entire synthesis and orchestration pipeline runs in pure air-gapped offline environments with zero network packets emitted.

Zoth binds each service strictly to the kernel loopback interface (127.0.0.1), ensuring that other machines on your local Wi-Fi or local area network cannot access your operator session or secrets unless you explicitly tunnel over an authenticated E2EE bridge such as Tailscale or WireGuard.

Never. Zoth Studio contains zero analytics trackers, zero telemetry probes, and zero phone-home routines. If you configure commercial BYOK model providers (such as Anthropic Claude, Google Gemini, or xAI Grok), requests travel directly from your client to the respective provider endpoint with no middleman server.

Your API keys are stored exclusively in operator memory or inside the encrypted zoth-vault-daemon. Keys are protected with Argon2id password-based key derivation (64 MB memory cost, 3 iterations) and XChaCha20-Poly1305 authenticated encryption. Keys are never written in plaintext to persistent log files or public repositories.

The Vault Daemon is a lightweight, zero-knowledge background process written in Rust/Python that provides encrypted credential storage for your workspace. It exposes loopback REST endpoints (POST /v1/vault/derive, POST /v1/vault/encrypt, POST /v1/vault/decrypt) allowing agents to request temporary authorization without exposing master keys.

Zoth features an integrated SimpleX Chat bridge on :8767 connecting to a headless SimpleX daemon on :5225. Because SimpleX has no user identifiers or central servers, operators can dispatch commands and receive agent telemetry from mobile devices over a metadata-free, post-quantum end-to-end encrypted queue.

Rather than relying on a single generic AI prompt, Zoth organizes 21 purpose-built agent archetypes — including Master Azoth (Lead Architect), Athena (Project Planner), Antigravity (Automation Engineer), Hermes (Deep Researcher), Draco (Security Reviewer), and Kitsune (3D & Visual Designer). Each agent has tailored tool capability grants, safety invariants, and persona memory.

Agents communicate over a shared on-disk file bus (agent-comms/) and WebSocket event mirror (:8989). For critical architectural decisions, Zoth executes a 3-Agent Consensus Protocol where independent models (e.g. Claude 3.5 Sonnet, Gemini 1.5 Pro, and Grok 2) generate parallel AST proposals, measure Shannon entropy differences, and synthesize a single hardened output.

Zoth features a universal Model Foundry supporting:

• Local Offline: Ollama (qwen2.5-coder, llama3.3, deepseek-r1, smollm2:360m) on :11434
• Frontier BYOK APIs: Anthropic Claude, Google Gemini, OpenAI GPT-4o, xAI Grok, DeepSeek, and Groq Cloud high-speed inference
• In-Browser WebGPU: Lightweight 360M parameter neural transformers running client-side with zero dependencies

Yes. Any script, binary, or CLI tool can be registered into Zoth’s tool registry (registry.local.json). You define a JSON Schema contract specifying input arguments, execution permissions, and expected return types, and all 21 agents instantly gain the ability to chain and invoke it.

Minimum Requirements (Cloud API / Lightweight Local): 4-Core CPU, 8 GB RAM, 2 GB disk space. Works smoothly on Intel/AMD PCs, Apple Silicon M1/M2/M3/M4, and Raspberry Pi 5.

Recommended (Full Local Swarm & GPU Inference): 8+ Core CPU, 32 GB RAM, NVIDIA RTX 3080/4080/4090 (16-24 GB VRAM) or Apple M-Series with 36GB+ Unified Memory.

Installation takes less than 60 seconds with our native packages and scripts:

• Linux (.deb / AppImage): Download zoth-studio_2.6.0_all.deb or run sudo dpkg -i zoth-studio*.deb
• Universal Shell Script: Run curl -fsSL https://zoth.nullai.tech/install.sh | bash
• Windows: Download the pre-packaged zoth-studio-windows.zip or run the PowerShell launcher .\install.ps1

The standard Zoth port topology is:

• :8088 — Web Studio Hub & Documentation (Static)
• :8484 — Operator Cockpit & Swarm Bus WebSocket (Private Loopback)
• :8787 — Argon2id BYOK Vault Daemon (Private Loopback)
• :11434 — Ollama Local Neural Inference Engine
• :5225 / :8767 — SimpleX Headless Daemon & Web Bridge

Yes. Zoth Studio is 100% free and open source released under the liberal MIT License. You are free to inspect the code, modify the agents, deploy on your hardware, and integrate it into your workflows without paying seat licenses or software royalties.

No. The Operator OS is $0. There are no user tiers, no feature lockouts, no time-limited trial periods, and no token markups. You bring your own hardware and your own keys.

Founder Patron is optional community sponsorship on GitHub. Patron contributions fund maintenance for the open forge: cutting and signing Debian/Windows packages, auditing security tools, and publishing free architectural whitepapers. Patron funding never locks or gates studio features.

This product is Zoth Studio. NullAI (nullai.tech) is our parent professional services and consulting firm providing enterprise SLAs, custom hardware companion fabrication, and bespoke multi-agent foundry pipelines for organizations that require dedicated engineering contracts.