opencode

opencode is a terminal-based coding agent. It reads an OpenAI-compatible provider from its config, so — like Hermes and OpenClaw — it's a first-class connect target: the proxy writes the config, and seeds it with the live model catalog, for you.

Before you start

Do the shared setup from How-to guides first: the proxy must be running (zs-proxy proxy start) and funded (zs-proxy fund).

The quick way: zs-proxy connect opencode

zs-proxy connect opencode                          # configure opencode to use the proxy
zs-proxy connect opencode --model qwen3-coder      # ...and set it as the default model
zs-proxy connect opencode --print                  # preview the merged config, write nothing

This merges a zerosignal provider into your existing ~/.config/opencode/opencode.jsonc (backing the original up to opencode.jsonc.bak) using opencode's @ai-sdk/openai-compatible adapter. It's JSONC, so your comments and other settings are preserved. What it writes:

{
  "provider": {
    "zerosignal": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "ZeroSignal",
      "options": {
        "baseURL": "http://127.0.0.1:9376/v1",   // your live proxy endpoint
        "apiKey": "zerosignal-local"             // placeholder — the proxy ignores it
      },
      "models": {
        "qwen3-coder": {
          "name": "qwen3-coder",
          "limit": { "context": 262144, "output": 65536 },
          "cost": { "input": 0.153, "output": 0.612, "cache_read": 0.0153, "cache_write": 0.153 }
        }
        // ...one entry per model in the live catalog
      }
    }
  },
  "model": "zerosignal/qwen3-coder"   // only when you pass --model
}

info

opencode requires each model to declare both a context and an output limit, so connect seeds the live catalog with each model's limit.context and limit.output. When a model doesn't advertise a max output, connect falls back to a sane budget (min(8000, context/4)) — the proxy still meters and bills on actual usage, so this only affects opencode's own client-side reservation.

info

Pricing. connect also seeds each priced model with a cost block (USD per 1M tokens) so opencode's session-cost readout reflects real spend instead of $0. The figures are the network's advertised rates grossed up by the protocol fee — i.e. what you actually pay — taken from the cheapest operator serving each model ("from $X"). Treat them as an estimate: the authoritative charge is the amount the chosen operator signs on-chain, which connect can't know in advance, and routing may land on a different-priced operator. cache_write mirrors the input rate (ZeroSignal has no cache-write premium); free or unpriced models get no cost block.

If opencode's config lives somewhere non-standard, point connect at it with OPENCODE_CONFIG=/path/to/opencode.jsonc (or --config-path).

Select the model and run

Start opencode and pick a zerosignal/… model — type /models to open the picker, or it'll use the default if you set one with --model:

opencode

Send a message; the first reply confirms the whole path — encrypt, pay, route — is working.

Troubleshooting

Symptom	Fix
connect says opencode wasn't detected	Name it explicitly — zs-proxy connect opencode configures the named tool even without detection, writing ~/.config/opencode/opencode.jsonc.
opencode isn't using the proxy	Pick a zerosignal/<model> in the /models picker, and confirm the zerosignal provider exists under provider in the config.
Connection refused / can't reach the endpoint	The proxy isn't running or the base URL is wrong. Confirm with curl http://localhost:9376/healthz and start it with zs-proxy proxy start.
wallet_unfunded / payment errors	Add funds with zs-proxy fund — see Wallet & funding.

What's next

• Connecting AI tools — every connect flag and the other supported tools.
• Routing preferences — pin an operator or set a price ceiling.
• Pricing — what an agent turn costs.
• How-to guides — connect another app.