Ollama Permissions Console

A local, two-pane web console that puts a permission layer in front of Ollama. Every capability the model reaches for is checked against what you set: allow, ask or deny. Everything stays on your disk.

Runs locally, PythonModels Ollama, tool-callingLicence CC BY-NC 4.0

Get it on GitHub Patreon

$ permissions
  internet_access    ask
  filesystem_access  ask
  image_generation   ask
  image_input        allow
  memory             always
› fetch the weather for Bouse, AZ
  ⏸ web_fetch wants the internet: allow?

THE PERMISSIONS PANE

How it works

You decide what the model may do

Permissions

Allow, ask or deny

Capability	Tools it gates	Starts as
Internet	web fetch	ask
Your files	read a file, list a folder	ask
Image generation	make an image (ComfyUI)	ask
Designs	generate a design	allow
Image input	attach an image	allow

On ask, the chat pauses until you approve or deny, and a waiting request survives a page refresh. Memory is always on, with a size cap.

Characters

Personalities with memory

Personalities and roles, with import and export as JSON
Lore, a scene note, and an automatic character-state summary
Long-term memory, kept in SQLite: global, per personality and per chat
Idle check-ins, from 5 minutes to an hour

Voice and images

Talk to it, show it things

Voice out: Supertonic 3 on the CPU, 10 stock voices plus your own
Voice in: faster-whisper on the CPU
Images: through ComfyUI, if you have it (optional)
See the exact prompt for each turn

Get started

Install

You need

Requirements

Python 3.11 or newer
Ollama, with a model that supports tool calling, like qwen3-vl, llama3.1 or mistral-nemo
ComfyUI, only for making images
No GPU needed for voice
The start and stop scripts are for Windows; the server itself is plain Python

Steps

Up and running

Clone the repository from GitHub.
In permissions-console/llm-permissions-app: pip install -r requirements.txt
ollama pull qwen3-vl
Run startup.bat and open http://localhost:8010.

The voice model downloads itself the first time it speaks.

Free for non-commercial use

CC BY-NC 4.0. Get it on GitHub.

GitHub