AE Agent
Using local models (Ollama / LM Studio)
Run an AI model entirely on your own machine — free, private, and offline. AE Agent connects to Ollama, LM Studio, or any OpenAI-compatible endpoint running locally. Nothing leaves your computer.
Local models are great for privacy and zero cost, but tool-use quality varies by model. For Agent mode, use a capable recent model —
llama3.1 or newer, or qwen2.5 or newer. Smaller or older models may struggle to call tools reliably.Ollama
1
Install Ollama
Download and install Ollama for your OS from ollama.com/download. It runs a local server at
http://localhost:11434.2
Pull and run a model
From a terminal, pull a tool-capable model and start it:The first run downloads the model. After that it stays available locally.
ollama run llama3.1
3
Connect it in AE Agent
Open Settings → Local AI, set the URL to:Click Detect models. AE Agent lists every model you've pulled. Pick one — it appears in the model picker under Local.
http://localhost:11434
LM Studio
1
Install and load a model
Install LM Studio from lmstudio.ai, download a model from its library, and load it.
2
Start the local server
In LM Studio, open the Local Server tab and start it. It exposes an OpenAI-compatible endpoint at
http://localhost:1234.3
Connect it in AE Agent
In Settings → Local AI, set the URL to:Click Detect models and select your loaded model.
http://localhost:1234
Custom OpenAI-compatible endpoints
Any server that speaks the OpenAI chat-completions API works the same way. Enter its base URL in Settings → Local AI and click Detect models. This covers self-hosted runtimes like llama.cpp's server, vLLM, and similar.
Default endpoints
| Ollama | http://localhost:11434 |
| LM Studio | http://localhost:1234 |
| Custom | Your OpenAI-compatible base URL |
If Detect models finds nothing, confirm the local server is actually running and the URL/port matches. A model also has to be pulled (Ollama) or loaded (LM Studio) before it shows up.