Selftuts

Where you learn on your own

AI

How to Connect OpenCode to Your Local Ollama Models

If you already use Ollama to run AI models locally, you can connect those models to OpenCode and use them for agentic coding. The best part is that you do not need to configure Ollama separately for every project. OpenCode supports a global configuration, which means you can configure Ollama once and then use your local models from OpenCode in any project.

In this tutorial, we’ll connect OpenCode to a local Ollama model using a global configuration. For this example we will use qwen3-coder:30b but you can use any compatible model available in your Ollama installation.

Bash
qwen3-coder:30b

Prerequisites

Before continuing, make sure you already have:

  • Ollama installed
  • OpenCode installed
  • At least one model downloaded in Ollama

How the Setup Works

There are three important components here.

  • OpenCode provides the coding-agent environment and tools.
  • Ollama runs the LLM locally and exposes an API that applications can use.
  • Qwen3-Coder is the actual coding model being executed by Ollama.

Once OpenCode knows where Ollama is running and which model to use, it can communicate with your local model.

Verify Ollama is Running

Before connecting OpenCode, let’s verify that the Ollama API is available. you need to run the below command from the terminal where Ollama is running

Bash
curl.exe http://127.0.0.1:11434/api/tags

You should receive JSON containing your installed models. If you can see your model here, Ollama is running correctly and its local API is accessible.

Bash
{
  "models": [
    {
      "name": "qwen3-coder:30b",
      "model": "qwen3-coder:30b",
      "modified_at": "2026-09-09T09:11:56.599224+05:30",
      "size": 18556700761,
      "digest": "06c1097efce0431c2045fe7b2e5108366e43bee1b4603a7aded8f21689e90bca",
      "details": {
        "parent_model": "",
        "format": "gguf",
        "family": "qwen3moe",
        "families": [
          "qwen3moe"
        ],
        "parameter_size": "30.5B",
        "quantization_level": "Q4_K_M",
        "context_length": 262144,
        "embedding_length": 2048
      },
      "capabilities": [
        "completion",
        "tools"
      ]
    },
    {
      "name": "qwen3:4b",
      "model": "qwen3:4b",
      "modified_at": "2026-03-02T20:15:30.6406779+05:30",
      "size": 2497293931,
      "digest": "359d7dd4bcdab3d86b87d73ac27966f4dbb9f5efdfcc75d34a8764a09474fae7",
      "details": {
        "parent_model": "",
        "format": "gguf",
        "family": "qwen3",
        "families": [
          "qwen3"
        ],
        "parameter_size": "4.0B",
        "quantization_level": "Q4_K_M",
        "context_length": 262144,
        "embedding_length": 2560
      },
      "capabilities": [
        "completion",
        "tools",
        "thinking"
      ]
    }
  ]
}

Create the Global OpenCode Configuration

You could create an opencode.json inside each individual project. But if you want to use Ollama across multiple projects, that quickly becomes repetitive. Instead, configure it globally once.

OpenCode’s global configuration location is:

Bash
~/.config/opencode/opencode.json

On Windows, this will normally correspond to:

Bash
C:\Users\<YOUR_USERNAME>\.config\opencode\opencode.json

You may need to create the directories if they don’t already exist.

Configure Ollama as an OpenCode Provider

Put the following inside the global opencode.json

Bash
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "ollama": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Ollama (local)",
      "options": {
        "baseURL": "http://127.0.0.1:11434/v1"
      },
      "models": {
        "qwen3-coder:30b": {
          "name": "Qwen3 Coder 30B"
        }
      }
    }
  },
  "model": "ollama/qwen3-coder:30b"
}

This small configuration is essentially the bridge between OpenCode and your locally running Ollama server.

Understanding the Configuration

Schema

Bash
"$schema": "https://opencode.ai/config.json"

The schema tells editors and tooling that this file follows the OpenCode configuration format. It helps with things such as validation and autocomplete.

Provider Configuration

The provider section is the most important part of this configuration.

Bash
"provider": {
  "ollama": {
    "npm": "@ai-sdk/openai-compatible",
    "name": "Ollama (local)",
    "options": {
      "baseURL": "http://127.0.0.1:11434/v1"
    },
    "models": {
      "qwen3-coder:30b": {
        "name": "Qwen3 Coder 30B"
      }
    }
  }
}

Whenever we define a provider in OpenCode, we essentially need to tell OpenCode three things:

Which provider are we using?

Here we are defining a provider called:

Bash
ollama

and giving it the display name:

Bash
Ollama (local)

How should OpenCode communicate with that provider?

Bash
"npm": "@ai-sdk/openai-compatible"

This tells OpenCode to use an OpenAI-compatible adapter to communicate with Ollama. It does not mean that we’re using OpenAI models. Ollama supports an OpenAI-compatible API, allowing OpenCode to communicate with our local Ollama server through the same API style.

Where is the provider running, and which models does it provide?

Bash
"options": {
  "baseURL": "http://127.0.0.1:11434/v1"
}

The baseURL tells OpenCode where the Ollama server is available. Finally, we list the models that we want to make available through this provider:

Bash
"models": {
  "qwen3-coder:30b": {
    "name": "Qwen3 Coder 30B"
  }
}

So the entire provider configuration can be understood as:

Bash
Provider
│
├── Name
│   └── Ollama
│
├── Adapter
│   └── OpenAI Compatible
│
├── Base URL
│   └── http://127.0.0.1:11434/v1
│
└── Models
    └── qwen3-coder:30b

If you have multiple Ollama models, you can list them all under the models section.

Default Model

After defining the provider and its available models, we tell OpenCode which model should be used by default:

Bash
"model": "ollama/qwen3-coder:30b"

OpenCode uses the format:

Bash
provider/model

Therefore:

Bash
ollama/qwen3-coder:30b
   │          │
Provider    Model

so whenever you will open the open code it will use the qwen3-coder:30b model from the Ollama provider by default.

Start OpenCode on any project and validate the Ollama integration

Start the OpenCode inside your project and see the details listed in the OpenCode screen

Bash
opencode

Test the Local Model

Now let’s verify that the integration actually works. Inside a test project, ask OpenCode:

Bash
Create a file called hello.py that prints
"Hello from my local AI coding agent"

If everything is configured correctly, OpenCode should invoke your local model and create the file. You can see in the below video that the hello.py file was created.

Conclusion

Connecting OpenCode to Ollama does not require adding configuration to every repository. You can do it once in the global configuration. The will help you to setup your own local agentic development.

Related Post

Embeddings In AI Models

Raj Ranjan

What is Transfer Learning in AI Models

Raj Ranjan

Neural Network explained in Simple Words

Raj Ranjan

Leave a Reply

Your email address will not be published. Required fields are marked *

You Missed

AI

How to Connect OpenCode to Your Local Ollama Models

AI

Embeddings In AI Models

Artificial Intelligence

Tokenisation in AI Models

AI Artificial Intelligence

What is Transfer Learning in AI Models