OpenCode on Seedrise

OpenCode settings at a glance

Checked against OpenCode's provider and config docs on 2 November 2026.

Why run OpenCode through Seedrise

Setup in five minutes

1. Install OpenCode. Follow OpenCode's install guide for macOS, Linux or Windows. Nothing Seedrise-specific here.
2. Get your Seedrise key. Request a rate card, receive your console login, and create a key for OpenCode in the console. Limit it to the models you plan to use and set a monthly budget.
3. Put the key in your environment. Add it to your shell profile (~/.zshrc or ~/.bashrc). The config reads it with {env:SEEDRISE_API_KEY}, so the key never sits in a file you might commit.
   bashCopy

   export SEEDRISE_API_KEY="sk-..."   # your Seedrise key

4. Add Seedrise to opencode.json. Use ~/.config/opencode/opencode.json for yourself, or opencode.json in the project root to share it through the repo. Model IDs have no vendor prefix.
   opencode.jsonCopy

   {
     "$schema": "https://opencode.ai/config.json",
     "provider": {
       "seedrise": {
         "npm": "@ai-sdk/openai-compatible",
         "name": "Seedrise",
         "options": {
           "baseURL": "https://api.seedrise.ai/v1",
           "apiKey": "{env:SEEDRISE_API_KEY}"
         },
         "models": {
           "gpt-6.1-sol": { "name": "GPT-6.1 Sol", "limit": { "context": 1050000, "output": 128000 } },
           "deepseek-v4-pro": { "name": "DeepSeek V4 Pro", "limit": { "context": 1000000, "output": 384000 } },
           "kimi-k3.1-pro": { "name": "Kimi K3.1 Pro" },
           "qwen3.8-max": { "name": "Qwen3.8 Max", "limit": { "context": 1000000, "output": 131072 } }
         }
       },
       "anthropic": {
         "options": {
           "baseURL": "https://api.seedrise.ai/v1",
           "apiKey": "{env:SEEDRISE_API_KEY}"
         },
         "whitelist": ["claude-opus-5-5", "claude-sonnet-5-5", "claude-haiku-5-5"]
       }
     },
     "model": "anthropic/claude-sonnet-5-5",
     "small_model": "anthropic/claude-haiku-5-5"
   }

5. Start and verify. Run opencode models seedrise to list the models from your config, then start opencode, choose a model with /models and send a first prompt.

What the config does

• seedrise uses @ai-sdk/openai-compatible, which calls /v1/chat/completions and works for every model on your key.
• anthropic keeps OpenCode's built-in Claude provider, with its model metadata, and sends it to Seedrise. The base URL ends in /v1 because this provider adds /messages itself. Claude Code is different: it takes the root URL.
• whitelist limits the Claude list to the models your key can call. Add claude-fable-5-1 only if it is enabled on your key.
• limit gives OpenCode each model's context window and maximum output, so it can show how much context is left. It is optional; the figures come from the model library.
• Prefer one provider for everything? Add the Claude IDs under seedrise.models and remove the anthropic block; Claude then runs over Chat Completions.

Models in this config

Provider list prices in USD per 1M tokens, input / output, as at 1 November 2026. Context windows checked 1 November 2026. Your Seedrise rate card is quoted per account against these prices.

Any other ID in the model library goes under seedrise.models the same way. See plans and pricing for how rates are set.

Check it is working

• opencode models seedrise lists the four Seedrise models as seedrise/<id>.
• In OpenCode, /models shows the Seedrise provider and only the Claude models in your whitelist.
• opencode run -m seedrise/deepseek-v4-pro "Say hello" answers without opening the interface, which is handy in scripts and CI.
• Usage appears in the Seedrise console within a minute, by key and model.

Known errors

Still stuck? Email support@seedrise.ai with the time of the request, the model ID and the key name.

For teams

Commit opencode.json to the repo so everyone gets the same providers and models, while each developer keeps their own key in SEEDRISE_API_KEY. OpenCode merges config files, so a short file like this can limit OpenCode to the Seedrise providers:

{
  "$schema": "https://opencode.ai/config.json",
  "enabled_providers": ["seedrise", "anthropic"],
  "model": "anthropic/claude-sonnet-5-5",
  "small_model": "anthropic/claude-haiku-5-5"
}

• To enforce it on company machines, put the config in OpenCode's managed directory, which users cannot override: /Library/Application Support/opencode/ on macOS, /etc/opencode/ on Linux, %ProgramData%\opencode on Windows.
• Give each team or project its own key, so spend shows up per project on the invoice.
• Set a monthly budget on every key. When a key reaches its budget, requests return 429 and the agent stops instead of running on.
• small_model handles session titles and other light tasks. Haiku 5.5 keeps that cost low.

OpenCode questions

Sources

• Custom providers, base URL, whitelist and troubleshooting: opencode.ai/docs/providers, checked 2026-11-02.
• Config locations, merging, managed settings, enabled_providers and {env:...}: opencode.ai/docs/config, checked 2026-11-02.
• opencode models and opencode run: opencode.ai/docs/cli, checked 2026-11-02.