Run AI models completely free on your own computer — no API keys, no usage costs, full privacy.
What is Ollama?
Ollama is a free, open-source tool that lets you run powerful AI models locally on your Windows, Mac, or Linux machine. Your data never leaves your computer.
Benefits:
- Completely free — no API costs
- Private — data stays on your machine
- Works offline — no internet required for inference
- Supports 100+ models including Llama, Mistral, Qwen, DeepSeek, and more
Step 1 — Install Ollama
Download and install Ollama for your operating system:
- Windows / Mac: Download from ollama.ai
- Linux: Run
curl -fsSL https://ollama.ai/install.sh | sh
After installing, Ollama runs as a background service on port 11434.
Step 2 — Download a Model
Open a terminal and pull a model. We recommend starting with a small, fast model:
ollama pull qwen2.5:0.5bRecommended models by hardware:
| Model | Size | RAM Required | Speed |
|---|---|---|---|
| qwen2.5:0.5b | 394MB | 2GB | Very Fast |
| llama3.2:latest | 2GB | 4GB | Fast |
| llama3:latest | 4.3GB | 8GB | Medium |
| qwen2.5:14b | 9GB | 16GB | Slow on CPU |
To list your installed models: ollama list
Step 3 — Download the AIDA Launcher
The one-click launcher starts Ollama and creates a secure tunnel so AIDA can reach your local model from anywhere.
Download the launcher for your OS:
- Windows: Download start-aida-local.bat (inside the AIDA app)
- Mac / Linux: Download start-aida-local.sh (inside the AIDA app)
You also need cloudflared in the same folder as the launcher:
- Windows: Download cloudflared.exe — rename to
cloudflared - Mac: Run
brew install cloudflared - Linux: Run
wget https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64 && chmod +x cloudflared-linux-amd64 && mv cloudflared-linux-amd64 cloudflared
Step 4 — Run the Launcher
- Windows: Double-click
start-aida-local.bat - Mac/Linux: Double-click
start-aida-local.shor run it from terminal
The launcher starts Ollama and creates a secure tunnel automatically. A public URL like https://xyz.trycloudflare.com will appear in the window — copy it.
Step 5 — Connect AIDA to Ollama
- Open AIDA → Settings
- Under LLM Provider, click Ollama (Local)
- Paste the tunnel URL in the Server URL field
- Click Save URL
- Click Auto-Detect Models
- Select your model from the list
- Start chatting!
Tools for Local Models
By default, local models run in chat-only mode — no calendar, email, or web tools. This keeps small models fast and reliable. You can change this with the Tools for local models setting on the provider card in Settings:
| Tier | What it does | Recommended model |
|---|---|---|
| None (default) | Chat + knowledge base only. Fastest. | Any, including 0.5B-1B |
| Curated | Lookups (time, weather, web search, memory, notes, reminders, calendar, email) + saving notes and memories | llama3.2:3b or larger |
| Full | Every AIDA tool | 8B+, ideally 70B |
Important — raise the context window. Ollama defaults to a 4096-token context, which is too small once tools are enabled. Before starting Ollama, set:
OLLAMA_CONTEXT_LENGTH=8192(On Windows, set it as an environment variable; on Mac/Linux, run export OLLAMA_CONTEXT_LENGTH=8192 in the terminal before the launcher.)
Signs the context is too small: tools get ignored, answers cut off halfway, or 524 timeout errors. If that happens, switch back to None or use a larger context.
Tips
Keep the launcher running while using AIDA with local models. Closing it disconnects the tunnel.
New URL each session — the tunnel URL changes every time you restart the launcher. Update it in AIDA Settings each time.
Permanent URL — for a permanent URL, create a free Cloudflare account and set up a named tunnel.
Slow responses? Use a smaller model (0.5B or 1.5B). Larger models are slow on CPU. A dedicated GPU makes a significant difference.
Tool calling — off by default for speed. Enable it in Settings → Ollama card → Tools for local models (see the section above). Needs a 3B+ model and a larger context window; known to work well: llama3.2:3b, qwen2.5:7b, mistral.