opencode-openrouter

OpenCode + OpenRouter: One Coding Agent, Every Model

Setup and configs that accompany the CloudYeti video: plugging OpenCode (an open-source terminal coding agent) into OpenRouter (one API key → hundreds of models, with routing and fallback), so you can switch the model behind your coding workflow without rebuilding your setup.

Why this stack: most coding agents bundle the agent and a fixed model path. OpenCode is the agent shell; OpenRouter is the model layer. Together you get a curated model menu, provider fallback when one is rate-limited, and the freedom to send cheap tasks to a cheap model and hard tasks to a frontier model — the same idea whether you're solo or on a team.

Steps below follow the official OpenCode + OpenRouter docs (openrouter.ai/docs/use-cases/opencode). Use a fresh OpenRouter key and keep it private.

Setup

1. Install OpenCode

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

2. OpenRouter API key

Create a key at openrouter.ai/keys, add a few dollars of credits.

3. Connect (beginner path)

Start OpenCode inside a real project, then connect:

cd /path/to/your/project
opencode

Inside OpenCode:

/connect      # choose OpenRouter, paste your key (it's stored locally)
/models       # pick the model you want to use

That's it — OpenCode is no longer tied to one model.

4. The config file (the power feature)

For a repeatable, team-friendly setup, define a curated model menu instead of picking from hundreds each time. Copy opencode.json into your project root:

curl -fsSL https://raw.githubusercontent.com/ravsau/ai-tutorials/main/opencode-openrouter/opencode.json -o opencode.json

It's just this:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "openrouter": {
      "models": {
        "anthropic/claude-sonnet-latest": {},
        "google/gemini-flash-latest": {},
        "deepseek/deepseek-v4-flash": {},
        "qwen/qwen3.6-coder": {}
      }
    }
  }
}

A sensible 4-model menu:

Role	Model	When
Strong default	anthropic/claude-sonnet-latest	hard coding tasks
Cheap + fast	google/gemini-flash-latest	simple edits, boilerplate
Cheap open	deepseek/deepseek-v4-flash	budget runs (fractions of a cent)
Coder	qwen/qwen3.6-coder	code-specific tasks

Swap these for whatever fits your budget — check live prices and quantization per provider at openrouter.ai/api/v1/models/<model>/endpoints.

5. Provider routing & fallback

OpenRouter picks a provider per request and can fall back if one is down. To pin or order providers (e.g. force a specific quantization, or avoid a rate-limited one), add routing to the request — see OpenRouter provider routing. If a provider is rate-limited, fallback keeps your coding session alive instead of killing it.

Reproduce the video test

Run the same task on two models and compare time, cost, files changed, and quality. The task is in task.md. Switch models with /models between runs, and reset the repo (or use a fresh branch) so the comparison is clean.

Files

File	What
opencode.json	Curated 4-model OpenRouter menu — drop into your project root
task.md	The identical test task to give each model

Notes / gotchas

• Model IDs on OpenRouter are provider/model (e.g. deepseek/deepseek-v4-flash). The ~ prefix some docs show is optional shorthand; the plain slug works.
• OpenRouter's docs are agent-readable: openrouter.ai/llms.txt is a clean index, and appending .md to any docs page returns the markdown (e.g. openrouter.ai/docs/use-cases/opencode.md).
• Cost is per-request and provider-dependent; turn on cost display and watch openrouter.ai/activity to see the real spend per model.