• English
  • OpenCode

    An open-source terminal agent with multi-model switching; connects to ByteSeek via the OpenAI-compatible protocol.

    Install

    Install script
    npm
    Homebrew

    macOS / Linux

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

    Website: opencode.ai.

    Prerequisites

    • OpenCode installed (opencode --version prints a version)
    • An OpenAI-format group key from the API Keys page

    Configure (one click from the console)

    No hand-written config needed — the console generates an opencode.json complete with the model list for your key:

    1. Open the API keys page, and in the key row's Actions column click More → Use key;
    2. Switch the dialog to the OpenCode tab and copy the generated opencode.json in full;
    3. Write it to ~/.config/opencode/opencode.json (opencode.jsonc also works), creating the file if it doesn't exist.
    SCREENSHOTUse key dialog, OpenCode tab
    The "Use key" dialog on the OpenCode tab: one opencode.json with the full model list and reasoning-effort variants — copy the whole thing (key redacted)

    The generated config points the default openai provider at ByteSeek, with context limits and reasoning-effort variants pre-filled per model. Structure excerpt (the dialog's content is authoritative — copy it wholesale):

    {
      "provider": {
        "openai": {
          "options": {
            "baseURL": "https://tokens.byteseek.ai/v1",
            "apiKey": "sk-your-key"
          },
          "npm": "@ai-sdk/openai",
          "models": {
            "gpt-5.6-sol": {
              "name": "GPT-5.6 Sol",
              "limit": { "context": 1050000, "output": 128000 },
              "options": { "store": false },
              "variants": { "low": {}, "medium": {}, "high": {}, "xhigh": {}, "max": {} },
              "tool_call": true
            }
          }
        }
      },
      "agent": {
        "build": { "options": { "store": false } },
        "plan": { "options": { "store": false } }
      },
      "$schema": "https://opencode.ai/config.json"
    }

    Extra notes (also flagged in the dialog):

    • The API key can stay out of the file — configure it with the client's /connect command instead;
    • You may also use a custom provider_id (such as byteseek) instead of the default openai provider, adding or removing models as needed;
    • Model IDs follow the model plaza; with a composite key, include the prefix.
    What you should see
    • Start opencode and press Ctrl+X to switch models — the list shows the provider from your config (default openai, or your custom ID) and its models;
    • The bottom-right corner shows the current model name;
    • A test message returns normally with no auth errors.

    Verify

    Start OpenCode, switch to a ByteSeek model in the picker and send a message. Charges appear on the usage page.

    Common issues

    SymptomFix
    401Check the apiKey field is complete
    Model missing from the listNot registered under models, or not in the key's group
    403Group / model / format mismatch, see Error Codes
    © 2026 ByteSeek Limited. All rights reserved.TermsPrivacyDisclaimer