opencode-info.json

{
  "productId": "opencode-multi-delegate",
  "type": "plugin",
  "displayName": "OpenCode Multi-Delegate",
  "description": "Dispatches any task to multiple AI models in parallel and produces a consolidated, cross-model analysis report. The plugin dynamically registers one analyst subagent per configured model, then an orchestrator skill coordinates them: all delegates run simultaneously, each returns structured findings with severity ratings, and the orchestrator synthesises a final report that marks each finding as confirmed, disputed, or rejected based on cross-model agreement. Useful for code review, architecture analysis, or any task where independent model perspectives improve reliability.",
  "repoUrl": "https://github.com/AlexMKX/opencode-multi-delegate",
  "homepageUrl": "https://github.com/AlexMKX/opencode-multi-delegate",
  "tags": ["multi-model", "delegation", "parallel", "analysis", "code-review", "subagents", "llm"],
  "installation": "## Installation\n\nInstall via npm (or any compatible package manager):\n\n```sh\nnpm install -g @alexmkx/opencode-multi-delegate\n```\n\nThen register the plugin in your `opencode.json`:\n\n```json\n{\n  \"plugin\": [\n    \"@alexmkx/opencode-multi-delegate\"\n  ]\n}\n```\n\n### Configure delegate models\n\nOn first run the plugin auto-creates `~/.config/opencode/multi-delegate.jsonc` with an empty template. Edit it to add the models you want to delegate to:\n\n```jsonc\n{\n  \"delegate_models\": [\n    { \"model\": \"anthropic/claude-opus-4-6\" },\n    { \"model\": \"openai/gpt-4o\" },\n    { \"model\": \"google/gemini-2.5-pro-preview\", \"variant\": \"medium\" }\n  ]\n}\n```\n\n`variant` is optional — omit it to use your provider's default settings.\n\nAfter editing the config, restart OpenCode. The plugin will register one analyst subagent per model (e.g. `@anthropic_claude-opus-4-6-delegate`) and write a `multi-delegate` skill to `~/.config/opencode/skills/multi-delegate/multi-delegate/SKILL.md` that the orchestrator can invoke.\n\n### Usage\n\nInvoke the `multi-delegate` skill (or just ask OpenCode to delegate your task to multiple models). The orchestrator dispatches all delegates in parallel, waits for all results, and returns a consolidated report with cross-model verdicts."
}