PlainClaw

PlainClaw is an OpenClaw plugin that forces jargon-heavy agent replies back into short, plain language.

It is the opposite of a "Humanizer" surface that tries to make AI sound more impressive. PlainClaw strips consultant tone, workplace buzzwords, and unexplained technical terms so users get the point in human words instead of being dazzled by fluff.

Why it is different

Most adjacent tools fall into one of three buckets:

• browser extensions that explain selected text after you highlight and send it

somewhere else

• humanizer tools that make AI output sound more polished or more human
• generic simplifiers that do not live inside the OpenClaw reply path

PlainClaw is different on purpose:

• OpenClaw-native: it works inside the reply pipeline instead of as copy/paste
• anti-buzzword: it removes fog instead of adding more style
• privacy-first: it rewrites locally and does not upload local chat content to a

plugin database

• confidence-aware: it pushes the model away from fake certainty and inflated

phrasing before the text is sent

• bilingual: it ships with both English and Chinese rewrite rules out of the box

What it does

• pushes the model toward plain, concise wording before generation
• rewrites final replies locally before they are written and sent
• matches the user's language and rough tone instead of sounding like generic AI
• supports instant /plainclaw on|off|status|adaptive|always commands
• does not upload local conversation data to any plugin database

Install

openclaw plugins install clawhub:@sheygoodbai/openclaw-plainclaw
openclaw plugins enable plainclaw

Then turn rewriting on in chat:

/plainclaw adaptive

openclaw plugins enable plainclaw only loads the plugin. The actual rewrite switch stays off until you run /plainclaw on or /plainclaw adaptive.

If ClawHub returns 429 Rate limit exceeded, use the source fallback:

git clone https://github.com/Sheygoodbai/plainclaw.git
cd plainclaw
openclaw plugins install -l .
openclaw plugins enable plainclaw

If the plugin does not show up immediately in openclaw plugins list, restart the gateway once after install or enable.

Commands

• /plainclaw on
• /plainclaw off
• /plainclaw adaptive
• /plainclaw always
• /plainclaw status
• /plainclaw help

Config

{
  "plugins": {
    "entries": {
      "plainclaw": {
        "config": {
          "enabled": true,
          "mode": "adaptive",
          "toneMatch": true,
          "jargonSensitivity": "balanced",
          "shortenOutput": true,
          "trackStats": true
        }
      }
    }
  }
}

Token note

PlainClaw reduces wasted output in the cheapest way OpenClaw currently allows:

• it tells the model up front to answer in plain language, which can reduce

output tokens at the source

• it also trims and rewrites the final reply locally without calling another

model

It cannot retroactively refund tokens that the upstream model already generated. It is designed to minimize waste, not pretend the platform supports impossible refunds.