OpenCode

OpenCode / SST. Docs checked 2026-10-12.

Terminal coding agent with first-party llama-server provider docs.

What it’s for

CLI agent that you point at llama-server via opencode.json.

Local endpoint

llama-server base: http://127.0.0.1:8080/v1
Strata base: http://127.0.0.1:8083/v1
Model name: provider.<id>.models keys (must match the server id)
API key: options.apiKey if the server requires one

Tool calling

Agent tools in the product. Use @ai-sdk/openai-compatible for /v1/chat/completions; @ai-sdk/openai if the model uses /v1/responses.

Long context

Per-model limit.context / limit.output in opencode.json. Example in their docs uses 128000 — set yours to what the server actually has.

ROCm / Radeon gotchas

No GPU stack. Same /v1 you already curled.

Hybrid local + cloud

Yes — cloud providers plus custom OpenAI-compatible.

This is the same question as the calculator’sfully localversushybridtoggle. This page does not quote cloud prices.

Official links

Sources

OpenCode — providers (llama.cpp example)

Copy-paste

More on the connect page. Prove http://127.0.0.1:8080/v1/models first.

OpenCode → llama-server :8080opencode.json

{
  "provider": {
    "llama": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "http://127.0.0.1:8080/v1"
      },
      "models": {
        "qwen": {
          "limit": { "context": 128000 }
        }
      }
    }
  }
}

Exact llama-server flags for our daily driver stay TODO in stacks.yaml. This is the port only. Docs example uses 128000 for limit.context — set yours to the server -c. Do not invent 262K here.

OpenCode → Strata :8083opencode.json

{
  "provider": {
    "llama": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "http://127.0.0.1:8083/v1"
      },
      "models": {
        "qwen": {
          "limit": { "context": 128000 }
        }
      }
    }
  }
}

Exact Strata flags for our rig stay TODO in stacks.yaml. This is the port only. Docs example uses 128000 for limit.context — set yours to the server -c. Do not invent 262K here.