Use Continuum’s hosted inference in OpenCode

What the provider block does

OpenCode treats providers as data. A key under provider in opencode.json names an adapter package, an endpoint, a credential, and a set of models; anything that speaks a shape one of its adapters understands can be added without OpenCode knowing the vendor exists. @ai-sdk/openai-compatible is the generic adapter for the OpenAI request shape, which is what Continuum serves at https://continuumcode.ai/v1.

Adding the provider does not remove anything. Whatever you had configured before — a subscription-backed provider, another API key, a local server — stays registered and stays in the picker, and Continuum joins the list rather than replacing it. That makes this a low-commitment change: you can switch models mid-session and compare.

What you need first

Setting OpenCode up this way requires Continuum Plus or higher. Inference keys do not mint on the Free tier, so the Create key button stays disabled until you subscribe, and the provider block below needs one to authenticate.

• OpenCode installed. Any recent version; the custom-provider mechanism is long-standing.
• A Continuum account on a paid plan. Personal API keys are a paid-plan feature: Plus at $25/mo, Max 100 at $100/mo, Max 200 at $200/mo, or Ultra at $500/mo as of August 2026. See pricing for what each includes.
• An editor and valid JSON. That is genuinely all.

Register Continuum in opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "continuum": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Continuum",
      "options": {
        "baseURL": "https://continuumcode.ai/v1",
        "apiKey": "cont_sk_..."
      },
      "models": {
        "claude-opus-5": { "name": "claude-opus-5" }
      }
    }
  }
}

"models": {
  "claude-opus-5":   { "name": "claude-opus-5" },
  "claude-sonnet-5": { "name": "claude-sonnet-5" },
  "gpt-5.6-sol":     { "name": "gpt-5.6-sol" },
  "grok-4.6":        { "name": "grok-4.6" },
  "gemini-3.5-flash": { "name": "gemini-3.5-flash" }
}

opencode models
opencode

The models map is the whole trick

Most OpenAI-compatible clients discover models: they call /v1/models, and whatever comes back is what you can pick. OpenCode does not work that way for a custom provider. The models map is the list, full stop. A model your plan serves, that curl confirms, that appears in every other client you have configured, will not appear in OpenCode unless it has an entry here.

These IDs were read from the live gateway on 17 August 2026. Each one needs its own entry in the map to be selectable.

The free lane

These models do not draw down your plan’s weekly inference allowance. Listing one in the map still needs a paid key, because that is what minting a key requires. They are chat-grade rather than agent-grade: useful for questions and small edits, and not what you want driving a long OpenCode run against a real repository. Note that these IDs contain slashes and a :free suffix, both of which must be reproduced exactly in the map key.

curl -s https://continuumcode.ai/v1/models \
  -H "Authorization: Bearer cont_sk_..."

When it does not work

On credential handling: unlike the Codex and Claude Code setups, which read a key from an environment variable, this one puts the key in a file in plaintext. Rotating means editing opencode.json, and the file deserves the same care as any other file holding a live credential — including keeping it out of a dotfiles repository you push somewhere.

Questions people ask

Keep reading