@qpthort/openclaw-plugin

Get your API key at neptrixai.com — you'll need one to use this plugin.

Neptrix governance plugin for OpenClaw agents (v3.0.0). Intercepts every tool call via the before_tool_call hook and evaluates it against constitutional safety rules before execution.

Does this send my data anywhere?

The plugin sends the tool name and parameters to Neptrix central for rule evaluation. Neptrix central stores governance signals only (decision type, rule ID, tool name). See the privacy section below for the full breakdown.

Install

openclaw plugins install @qpthort/openclaw-plugin

Configuration

Add your settings to ~/.openclaw/openclaw.json:

{
  "plugins": {
    "entries": {
      "@qpthort/openclaw-plugin": {
        "enabled": true,
        "config": {
          "apiKey": "YOUR_API_KEY",
          "serverUrl": "https://api.qpthort.com",
          "logDecisions": true
        }
      }
    }
  }
}

Get your API key at neptrixai.com.

| Option | Type | Default | Description | |---|---|---|---| | apiKey | string | "" | Your Neptrix API key (required) | | serverUrl | string | https://api.qpthort.com | Governance API endpoint | | logDecisions | boolean | true | Print decisions to console |

Restart the OpenClaw gateway after configuring. You should see:

[Neptrix] Active -> https://api.qpthort.com

Task Type Declaration

By default, the plugin declares task_type: "CODING", which loads H-series rules (block credentials, irreversible ops, dangerous commands). For non-coding agents, declare a task type in your agent config:

{
  "taskType": "OUTREACH"
}

Supported values:

| Task Type | Rule Set | Use Case | |---|---|---| | CODING | H-series | Development, devops, infrastructure | | OUTREACH | C-series | Email, social media, messaging | | COMMUNICATION | C-series | Same as OUTREACH | | RESEARCH | H-series (safe fallback) | Information gathering | | CONTENT_CREATION | H-series (safe fallback) | Writing, design, media |

The plugin reads taskType or task_type from event.agentConfig.

What Data Goes Where

| Data | Neptrix Central | |---|---| | Tool name | Yes | | Tool parameters | Yes (for rule evaluation) | | Governance decision | Yes | | Matched rule ID | Yes | | Confidence score | Yes | | Session ID | Yes | | Agent/model ID | Yes |

Neptrix central receives tool parameters for rule evaluation (it needs to inspect command text to detect credential access, destructive operations, etc.).

Fail-Secure Behaviour

The plugin blocks the tool call if any of the following occur:

• Neptrix API is unreachable (network error or timeout)
• Response is malformed (missing block boolean)
• Server returns a non-200 status code
• No API key is configured

The timeout is 5 seconds. The hook runs at priority 1000.

Privacy

Neptrix central processes tool calls to evaluate them against constitutional rules and returns an ALLOW or BLOCK decision. It stores governance signals (decision, rule ID, tool name, agent ID) for audit purposes.

Changelog

v3.0.0

• Added task_type declaration support -- non-coding agents get the right rule set
• Updated default API endpoint to https://api.qpthort.com
• Fail-secure on malformed responses (missing block field now triggers block)

License

ISC

Need an API key? Sign up at neptrixai.com to get started.