Ziva vs AI Assistant Hub: Paid Agent or Free Chat
May 25, 2026
Ziva and AI Assistant Hub by FlamxGames sit at the two ends of the Godot AI plugin spectrum. AI Assistant Hub is free, open source, and runs entirely on your machine if you point it at Ollama . Ziva is a paid agent with a managed runtime that acts on the editor directly. Both have valid use cases. Picking depends on whether you want to pay for capability or pay for control.
TL;DR
| If you want… | Pick |
|---|---|
| Free + open source + local LLM via Ollama | AI Assistant Hub |
| Agent that actually edits scenes and files | Ziva |
| Multiple simultaneous chat sessions with different model backends | AI Assistant Hub |
| Asset generation (sprites, 3D models) built in | Ziva |
| Bring your own API key (Gemini, OpenRouter) | AI Assistant Hub |
| Managed credits with frontier-model access included | Ziva |
| Save reusable quick prompts in the dock | AI Assistant Hub |
| Live debugger reading + editor screenshots | Ziva |
What AI Assistant Hub actually is
AI Assistant Hub is a Godot 4 plugin that adds a dock with one or more AI chat sessions. Each session connects to a backend: Ollama for local models (Llama, Mistral, codestral, anything you can pull), Gemini with a free API key, OpenRouter for paid access to many providers, or others. The plugin can read code you highlight in Godot’s script editor, write code into the editor (insert at cursor), and save reusable quick-prompts for common asks.
The plugin is MIT-licensed and installable through the Godot Asset Library. Free to use forever; the only cost is whatever you spend on API providers (zero if you stay on Ollama).
What Ziva is
Ziva is a Godot plugin with a built-in agent. The agent calls Godot’s EditorInterface directly to add nodes, configure properties, wire signals, edit TileMapLayer cells, generate sprites and 3D models, and read live debugger output. The plugin manages model selection (Claude Opus / Sonnet / Haiku, GPT-5, Gemini, Deepseek) and API credits as a single managed service.
Free tier (20 credits to try frontier models). $20/mo Pro for unlimited frontier-model usage.
Feature comparison
| Feature | Ziva | AI Assistant Hub |
|---|---|---|
| Primary purpose | Agent (changes things) | Chat (answers things) |
| Scene tree manipulation | Yes, direct API | No |
| Asset generation (sprites, 3D) | Built in | No |
| Live debugger reading | Yes | No |
| Editor screenshots for vision context | Yes | No |
| Multiple simultaneous chat sessions | One agent | Many chats |
| Reusable saved prompts | Per chat | First-class feature |
| Local LLM (Ollama, LM Studio) | No | Yes |
| Bring-your-own API key | No (managed) | Yes (Gemini, OpenRouter) |
| Pricing | Free (20 credits), $20/mo Pro | Free |
| Source | Closed | MIT open source |
| Model selection | Per task, in-dock | Per chat session |
| Highlight-and-ask | Yes | Yes |
Where AI Assistant Hub wins
Free forever, no API costs if you stay local. This is the headline. If you have a GPU that can run a quantized 13B model via Ollama, AI Assistant Hub gives you AI assistance with zero ongoing cost. For students, hobbyists, and anyone allergic to subscriptions, that matters.
Privacy by default. Run Ollama locally and your code never leaves your machine. No cloud provider sees your project. For proprietary or sensitive work, the local-first option is the safest.
Multiple simultaneous chats. Open three chats: one wired to a local Llama for general questions, one wired to Gemini Flash for fast code, one wired to OpenRouter pointing at Claude for hard logic. AI Assistant Hub lets you switch between them without restarting anything.
Reusable quick prompts. Save “convert this to typed GDScript” or “explain this node hierarchy” as a one-click button. Power-user feature for repetitive workflows.
Open source, MIT. You can fork it, audit it, modify it, and ship a custom version. For teams that need to know exactly what their tools do, source access is non-negotiable.
Where Ziva wins
It acts on the editor. AI Assistant Hub answers questions; Ziva does the work. If you ask Ziva to “build a state machine with Idle, Walk, Attack states for this enemy”, it instantiates the nodes, writes the script, wires the transitions, and saves. AI Assistant Hub would write a tutorial explaining how to do those steps; you would do them yourself. See what Ziva does in Godot for the API-level detail.
Asset generation. Sprites via Retrodiffusion , 3D models via a text-to-3D service, both writing into res:// with the correct .import config. AI Assistant Hub doesn’t generate assets.
Frontier models without managing API keys. With AI Assistant Hub you set up Ollama, manage API keys, manage billing across providers, and update model lists as new ones release. With Ziva you sign in and the right frontier model is available for the right task with one click. The $20/mo covers the model usage; you don’t manage credentials.
Live debugger reading. Run the game, hit a crash, Ziva already has the stack trace and proposes a fix. AI Assistant Hub requires you to paste error messages into chat.
Editor screenshots. Some bugs are visual (“this button is positioned wrong”). Ziva can capture an editor screenshot and pass it as image context. AI Assistant Hub is text-only.
When to pick each (honest)
Pick AI Assistant Hub if:
- You have a GPU that runs a local LLM well enough for your work
- You are allergic to subscriptions or working on a personal project
- You need open source for compliance reasons
- You prefer chat-style help over agent-style action
- You want to bring your own API keys and pay providers directly
Pick Ziva if:
- Your time is worth more than the $20/mo
- You want the AI to do the work, not explain how to do the work
- You need sprite or 3D asset generation in your workflow
- You want one tool covering scenes + code + assets + debugger
- You are shipping commercial work and want a managed runtime with documented privacy
Use both:
- Local AI Assistant Hub + Ollama for quick chat or explanations on sensitive code
- Ziva for the heavy lifting on the parts that aren’t sensitive
Related reading
- Best AI Tools for Godot in 2026: the 11-tool landscape
- What Ziva actually does in Godot: API-level capability walkthrough
- Ziva vs Summer Engine: plugin vs AI-native engine
- AI coding tools and code privacy: privacy posture overview