OpenClaw Advisor Plugin

Advisor adds an advisor tool and /advisor chat command to OpenClaw. The tool asks a configured LLM for a focused second opinion on a plan, implementation, review, or design tradeoff.

Use it when the agent needs another read before doing something risky or ambiguous:

• review a complex plan before execution
• check edge cases in a proposed implementation
• get a second opinion on requirements or architecture choices
• flag likely security issues, side effects, or unsafe assumptions

Install

openclaw plugins install clawhub:openclaw-advisor-plugin

Configuration

Enable the plugin and allow the advisor tool in OpenClaw config:

{
  "plugins": {
    "entries": {
      "advisor": {
        "enabled": true,
        "config": {
          "modelRef": "provider/model"
        },
        "llm": {
          "allowModelOverride": true
        }
      }
    }
  },
  "tools": {
    "alsoAllow": ["advisor"]
  }
}

Chat Commands

/advisor status

Shows the current advisor model. If no model is configured, advisor calls use the agent active model.

/advisor model <provider/model>

Sets the advisor model live. This writes plugins.entries.advisor.config.modelRef and enables plugins.entries.advisor.llm.allowModelOverride for the advisor plugin. The next advisor tool call uses the new model without restarting OpenClaw.

/advisor model clear

Clears the advisor model override. Future advisor calls use the agent active model.

Tool

advisor

• question - required question, problem, or decision for review.
• context - optional background, code snippets, constraints, or prior conclusions.

The advisor only sees the question and optional context passed into the tool call, so include the relevant code, constraints, and current conclusion in the request.