opencode-multi-ai-review

Reusable opencode plugin for running one branch code review across multiple selected models and returning one synthesized Markdown report.

Install from npm

Install the plugin into the current project's opencode config:

opencode plugin opencode-multi-ai-review

To install it globally instead:

opencode plugin --global opencode-multi-ai-review

You can also add it manually to your opencode config:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["opencode-multi-ai-review"]
}

Command registration is automatic. The runtime command template comes from src/command.ts; commands/multi-review.md is shipped as a packaged readable reference, so npm users do not need to copy a command file into each project.

Restart opencode after changing plugin or command configuration.

Install from a local clone

For local development from a clone, build this package and point opencode at the local plugin file:

cd /absolute/path/to/opencode-multi-ai-review
pnpm install
pnpm build

Then configure the plugin with an absolute file URL:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["file:///absolute/path/to/opencode-multi-ai-review/dist/index.js"]
}

Command registration is automatic for local development too. Rebuild the plugin after changing the runtime command template in src/command.ts or other TypeScript source. commands/multi-review.md is a packaged readable reference and should stay aligned with the runtime template.

Restart opencode after changing plugin or command configuration.

Use

Run the command in opencode:

/multi-review

The command runs opencode models, offers the previous model selection from .opencode/multi-ai-review/state.json when available, asks which models to use through the built-in question tool, then calls multi_ai_code_review.

The plugin writes .opencode/multi-ai-review/.gitignore next to the generated state so state.json stays local and is not accidentally committed.

To override the auto-detected base ref:

/multi-review origin/main

To add extra focus:

/multi-review origin/main focus on auth, data loss, and missing tests

Output

The final report includes:

• a findings table with severity, confidence, models, location, recommendation, and action;
• a Do Not Address Yet section for likely false positives;
• reviewer failures when a selected model fails or returns unparseable JSON;
• a partial-review notice when the diff is truncated.

Options

Plugin options can be provided with opencode's tuple plugin config form:

{
  "plugin": [
    ["opencode-multi-ai-review", {
      "maxDiffBytes": 200000,
      "maxDiffLines": 6000,
      "maxFiles": 200
    }]
  ]
}