OpenCode Setup

OpenCode is an open-source, single-binary AI coding agent with built-in ACP support — no adapter needed. It's the easiest way to run Agent Console fully offline against a local model.

Install and Configure

1. Install OpenCode with its one-line installer:

macOS/Linux

curl -fsSL https://opencode.ai/install | bash

Windows

:: See opencode.ai for the current Windows install options

2. Find the installation path:

macOS/Linux

which opencode
# Example output: /Users/you/.opencode/bin/opencode

Windows

where.exe opencode

3. Open Settings → Agent Console. Switch to the OpenCode section.

4. Configure:

   • Path: opencode (works in many cases via login shell PATH; or set the full path from step 2, or click Auto-detect)
   • Arguments: acp (set by default)

Choose a model

Agent Console only launches OpenCode over ACP — it doesn't pick the model. OpenCode chooses its own model and sign-in through its config, so there's no API key field in the OpenCode section here.

Pick a model in OpenCode itself: run opencode once in a terminal, use the /connect command to add a provider, then /models to choose one. OpenCode supports cloud providers (Anthropic, OpenAI, and many others) as well as local models.

Run local models (offline) with ollama

OpenCode can talk to a local model server, so your chats never leave your machine. This is configured in OpenCode's own config file, not in Agent Console.

1. Install ollama and pull a model with a large context window and strong tool-calling (for example, a Qwen-Coder variant):

   ollama pull qwen2.5-coder

2. Add ollama as a provider in 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": {
           "qwen2.5-coder": {
             "name": "Qwen2.5 Coder (local)",
             "limit": {
               "context": 65536,
               "output": 8192
             }
           }
         }
       }
     }
   }

3. Run opencode, then /models, and pick the local model.

TIP

Give the model plenty of context — agent and coding work wants a 64k+ window. Context length is set on the ollama server, not in Agent Console or OpenCode: ollama defaults to ~4k and ignores the value until you raise it at the server. On macOS run launchctl setenv OLLAMA_CONTEXT_LENGTH 65536 and restart Ollama (or add PARAMETER num_ctx 65536 to a Modelfile). The limit.context in opencode.json only tells OpenCode how much it may send — it can't raise the server's window.

INFO

Agent Console spawns opencode acp and speaks the protocol; the model backend lives entirely in OpenCode. That separation is deliberate — the plugin stays model-agnostic, so the same OpenCode setup works whether you point it at ollama, a cloud provider, or anything else OpenCode supports. Any environment variables you set in the OpenCode section apply to the OpenCode process only, not to the model backend — so a server-side setting like OLLAMA_CONTEXT_LENGTH belongs on the ollama server, not in that field.

Verify Setup

1. Click the robot icon in the ribbon or use the command palette: "Open chat"
2. Switch to OpenCode from the agent dropdown in the chat header
3. Try sending a message to verify the connection

Having issues? See Troubleshooting.