Aigon can mix local and cloud models in the same workflow. Implement a feature on your own hardware, then review it through a cloud provider — or vice versa. This guide walks through the setup using OpenCode with Ollama, but the same pattern works with any agent that supports an OpenAI-compatible endpoint (LM Studio, llama.cpp, vLLM, etc.).

Prerequisites

• Aigon installed and a project initialised (aigon init)
• OpenCode  installed (npm i -g opencode-ai) and the op agent configured (aigon install-agent op)
• A local model server — this guide uses Ollama 

1. Pull a local model

Pick a model that fits your hardware. As a rough guide:

ollama pull qwen3:4b

Start the server if it isn’t already running:

ollama serve

2. Configure OpenCode to use the local provider

OpenCode uses the Vercel AI SDK and supports any OpenAI-compatible endpoint via the @ai-sdk/openai-compatible adapter.

Add an Ollama provider to your OpenCode config at ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "ollama": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Ollama (local)",
      "options": {
        "baseURL": "http://localhost:11434/v1"
      },
      "models": {
        "qwen3:4b": {
          "name": "Qwen3 4B (local)",
          "limit": { "context": 32768, "output": 8192 }
        }
      }
    }
  }
}

OpenCode also needs an auth entry, even for local servers that don’t require one. Add a placeholder to ~/.local/share/opencode/auth.json:

{
  "ollama": { "type": "api", "key": "sk-local" }
}

3. Add the model to Aigon’s dashboard

Aigon’s model dropdowns are driven by each agent’s model options. To add your local model without editing any Aigon source files, use the customModelOptions key in your config.

For a per-user setting, add to ~/.aigon/config.json:

{
  "agents": {
    "op": {
      "customModelOptions": [
        {
          "value": "ollama/qwen3:4b",
          "label": "Qwen3 4B via Ollama (local)",
          "pricing": { "input": 0, "output": 0 }
        }
      ]
    }
  }
}

For a per-project setting, add the same block to .aigon/config.json in your repo.

Custom models appear at the top of the dropdown, above the shipped cloud options. Restart the dashboard to pick up the change:

aigon server restart

4. Test the connection

Verify OpenCode can reach your local model:

opencode --model ollama/qwen3:4b

You should see Build · Qwen3 4B (local) Ollama (local) in the OpenCode TUI. Type a simple prompt to confirm the model responds.

Mix local and cloud models

The real value is using different models for different steps. For example:

# Implement with your local model (free)
aigon feature-start <ID> op --model ollama/qwen3:4b
 
# Review with a cloud model (different perspective)
aigon feature-code-review <ID> op --model openrouter/qwen/qwen3-235b-a22b-07-25

Or select the models from the dashboard start modal — your local model appears in the model dropdown alongside the cloud options.

Adding more models

To add another local model, pull it with Ollama, add it to both configs:

1. Register it in OpenCode’s config (~/.config/opencode/opencode.json → provider.ollama.models)
2. Add it to Aigon’s config (customModelOptions array)

Multiple custom models are supported — they all appear at the top of the dropdown.

Other local model servers

The same pattern works with any OpenAI-compatible server. Replace the provider block in your OpenCode config:

LM Studio (default port 1234):

{
  "provider": {
    "lmstudio": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "LM Studio (local)",
      "options": { "baseURL": "http://127.0.0.1:1234/v1" },
      "models": {
        "your-model-id": { "name": "Your Model" }
      }
    }
  }
}

llama.cpp / llama-server (default port 8080):

{
  "provider": {
    "llama.cpp": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "llama-server (local)",
      "options": { "baseURL": "http://127.0.0.1:8080/v1" },
      "models": {
        "your-model-id": { "name": "Your Model" }
      }
    }
  }
}

Troubleshooting