OpenCode + TOKNADO

How to use OpenCode with TOKNADO (custom provider in opencode.json)

OpenCode is a terminal coding agent with a long list of built-in providers, plus a documented path for any OpenAI-compatible endpoint: declare the provider in opencode.json, give it a base URL, list the model IDs, and they appear in the model picker.

That makes TOKNADO a few lines of config. Here is the complete setup, including where to keep the key and the base URL mistake that causes most failures.

Start free trial Jump to setup

What you need

• OpenCode installed.
• Your TOKNADO endpoint and API key.
• At least one model ID from your TOKNADO dashboard.

Setup: OpenCode with TOKNADO

1. Set your API key

   Keep the key in the environment instead of the config file.

   export TOKNADO_API_KEY=YOUR_TOKNADO_API_KEY

2. Create opencode.json

   Use opencode.json in your project root for a per-project setup, or ~/.config/opencode/opencode.json for a global one. The provider ID (toknado) is your choice and becomes the first half of the model reference.

   {
     "$schema": "https://opencode.ai/config.json",
     "provider": {
       "toknado": {
         "npm": "@ai-sdk/openai-compatible",
         "name": "TOKNADO",
         "options": {
           "baseURL": "https://toknado.up.railway.app/v1",
           "apiKey": "{env:TOKNADO_API_KEY}"
         },
         "models": {
           "toknado/auto": { "name": "TOKNADO model" }
         }
       }
     },
     "model": "toknado/toknado/auto"
   }

3. Alternative: store the key with /connect

   Instead of the env var, run /connect inside OpenCode, scroll to Other, enter the provider ID (toknado), and paste the key.

4. Pick the model and run

   Start OpenCode and choose toknado/toknado/auto from the model list if it isn't already your default.

   opencode

Notes for real-world use

• OpenCode needs you to declare the models you want exposed. A compatible endpoint's model list doesn't always include context-window or output-limit metadata, so set limits per model if you hit truncation.
• The @ai-sdk/openai-compatible package uses the chat completions route. That matches what OpenAI-compatible gateways provide.
• Agents spend requests fast. Compare your usual session volume with your plan's rolling hourly limit.

Troubleshooting OpenCode + TOKNADO

FAQ