AI work that stays private,
routed, and under control.

A local-first AI coding workspace: run sensitive work on your machine, route heavier tasks only when approved, and keep every write behind a clear permission gate.

0required API keys for local work

17file, shell, and GitLab tools

Askbefore writes, shell, or cloud use

Controls

Private by default

Local Ollama models handle sensitive code without sending prompts to external APIs.

Cost-aware routing

Small tasks stay local. Cloud becomes an explicit escalation, not a hidden default.

Permissioned execution

Read-only actions run freely. Writes and shell commands stop for approval.

How the agent loop works

Classify the task

Decide local, tool-use, or cloud escalation.

Use sandboxed tools

Read files, search code, inspect GitLab.

Ask before risk

Writes, shell, and cloud all require approval.

Return the result

Final answer with full audit of what ran.

Run it locally

Clone, install, start Ollama, and open the web UI — no API key required.

quick-start · PowerShell, macOS, Linux

git clone https://github.com/sechan9999/LLMai.git
cd LLMai
pip install -e .

ollama serve
ollama pull qwen2.5-coder

vixcode-server
# open http://localhost:7777

Try the chat demo first

If Ollama is already running on your machine, the Chat tab connects directly via localhost:11434. Your prompts never leave the browser.

AI work that stays private,routed, and under control.