Automate SEO Meta Descriptions with Paradime, OpenClaw, and Google Sheets

TL;DR — Wire an OpenClaw agent to your Google Sheets content inventory, let it generate optimized meta descriptions page-by-page, schedule the job through Paradime Bolt, and sleep well knowing a weekly cron keeps your metadata fresh. This guide walks you through every step with a "time-to-first-clue" debugging mindset so you can reproduce the setup, ship fast, and fix faster.

What Is Paradime?

Paradime is an AI-native platform purpose-built for dbt™ practitioners. Think of it as Cursor for data: an all-in-one workspace that replaces dbt Cloud™ with a faster IDE, smarter scheduling, and built-in FinOps monitoring. Three pillars matter for this tutorial:

Bolt is the workhorse here. It runs arbitrary Python scripts alongside dbt™ commands, manages secrets as environment variables, and surfaces AI-powered debugging when something breaks. We will use Bolt to schedule our OpenClaw SEO script.

Why Paradime for this workflow? You already manage your data models in dbt™. By adding a Python script to the same repo and scheduling it through Bolt, your SEO automation lives alongside your data pipeline — one repo, one scheduler, one set of alerts.

What Is OpenClaw?

OpenClaw (formerly Moltbot / Clawdbot) is an open-source AI agent framework. It runs on your own device, connects to LLMs (Claude, GPT, local models via Ollama), and exposes a Python SDK you can use to orchestrate tasks programmatically.

For our use case the SDK is all we need:

Install with:

Verify the installation:

OpenClaw's agent.run() method sends a prompt, collects the response, and returns it — perfect for a loop over spreadsheet rows.

Architecture at a Glance

Figure 1 — End-to-end flow: Bolt triggers the script, the script reads from and writes to Google Sheets, OpenClaw generates the meta descriptions.

Setup: openclaw-sdk + Google Sheets API

Prerequisites

Step 1 — Create a Google Service Account

1. Go to Google Cloud Console → APIs & Services → Credentials.

2. Click Create Credentials → Service Account.

3. Assign the Editor role (or a custom role with Sheets read/write).

4. Under the new service account, click Keys → Add Key → JSON.

5. Download the JSON key file — this becomes your GOOGLE_CREDENTIALS_JSON.

6. Enable Google Sheets API and Google Drive API in the console.

7. Share your target spreadsheet with the service-account email (the client_email field in the JSON).

Step 2 — Get Your OpenClaw API Key

OpenClaw stores secrets via environment variables or its config file (~/.openclaw/openclaw.json). For Paradime Bolt, we only need the API key as an env var:

You can also store it in OpenClaw's config block:

Step 3 — Initialize the Project

Inside your dbt™ repo (the same repo Bolt watches), create a folder for the script:

Add dependencies to pyproject.toml:

The Script: Read, Generate, Write Back

Below is the full Python script (scripts/seo_meta_generator.py). It reads URLs and content from a Google Sheet, sends each row to OpenClaw for meta-description generation, and writes the result back.

How It Works — Step by Step

Figure 2 — Decision tree inside seo_meta_generator.py. The script skips rows that already have a meta description or lack content, minimizing API calls and cost.

Environment Variables

The script depends on two secrets. Here is how to set them in Paradime Bolt.

Adding Secrets to Paradime Bolt

1. Navigate to Settings → Workspaces → Environment Variables.

2. In the Bolt Schedules section, click Add New.

3. Enter the Key (e.g., OPENCLAW_API_KEY) and Value.

4. Click Save (💾 icon).

5. Repeat for GOOGLE_CREDENTIALS_JSON.

Bulk upload option: If you have many variables, prepare a CSV with Key,Value columns (no spaces) and use the Bulk Upload button.

Your Python script reads them with:

These variables are scoped to Bolt schedules only — they are not exposed in the Code IDE environment, keeping production secrets isolated from development.

Bolt Schedule: Cron Weekly or On-Demand

Option A — Schedules as Code (YAML)

Add the following to paradime_schedules.yml in your repo root:

Cron tip: Validate your expression at crontab.guru. Use 'OFF' to disable the schedule without deleting it.

Option B — On-Demand via Bolt API

Paradime's Bolt API lets you trigger any schedule programmatically:

This is ideal for event-driven runs — for example, triggering the script whenever a new page is added to the spreadsheet via a Zapier or Google Apps Script webhook.

How Bolt Processes the Schedule

Figure 3 — Bolt injects environment variables before executing commands in sequence. If poetry install fails, the pipeline stops and alerts fire immediately.

Monitoring and Debugging

Paradime Radar gives you four dashboards to monitor your pipeline:

DinoAI-Powered Debugging

When a Bolt run fails, Paradime's DinoAI analyzes the logs instantly and suggests fixes. Instead of scrolling through raw output, you get a structured diagnosis:

Setting Up Alerts

Configure alerts so failures reach you immediately:

For deeper observability, integrate with Elementary to get anomaly-based alerts.

Troubleshooting Common Issues

Use this decision tree when a run fails — it is designed for a "time to first clue" mindset.

Figure 4 — Incident-friendly decision tree. Start at "Bolt run failed" and follow the branches to your first clue.

Quick-Fix Reference Table

Reproducing Locally

Before pushing to Bolt, test locally:

If it works locally but fails in Bolt, the issue is almost always an environment variable that wasn't saved correctly.

Evaluating Meta Description Quality with dbt™-llm-evals

Once your meta descriptions are generated, you may want to evaluate their quality at scale. Paradime's open-source dbt™-llm-evals package lets you do exactly that — directly inside your data warehouse, with no data egress.

Quick Setup

Add the package to your dbt™ project:

Configure Evaluation Criteria

Build a dbt™ Model That Evaluates Your Meta Descriptions

Monitor Results

This closes the loop: generate → evaluate → improve.

Wrapping Up

Here's what you built:

Figure 5 — The complete stack: spreadsheet → script → OpenClaw → spreadsheet, orchestrated by Bolt, monitored by Radar, and quality-checked by dbt™-llm-evals.

Key Takeaways

1. Paradime Bolt is not just a dbt™ scheduler — it runs Python scripts with full dependency management (Poetry), secret injection, and SLA monitoring.

2. OpenClaw's Python SDK makes LLM-powered content generation as simple as agent.run(prompt).

3. Google Sheets serves as the lightweight content inventory — no database required for the SEO team.

4. Environment variables (OPENCLAW_API_KEY, GOOGLE_CREDENTIALS_JSON) are the only secrets you need. Store them once in Bolt; never hard-code them.

5. Schedules as code (paradime_schedules.yml) means your automation is version-controlled, reviewable, and reproducible.

6. dbt™-llm-evals lets you evaluate meta description quality warehouse-natively, closing the feedback loop.

Next Steps

• Scale up: Process thousands of pages by splitting into batches across multiple Bolt runs.

• Add keywords: Extend the spreadsheet with a "target keyword" column and inject it into the prompt.

• A/B test: Generate two variants per page and track CTR in Google Search Console.

• Integrate with CI: Use Bolt's merge trigger to regenerate descriptions whenever new pages land in main.

Further Reading:

• Paradime Bolt Documentation

• Paradime Environment Variables

• OpenClaw Getting Started

• OpenClaw Python SDK Guide

• dbt™-llm-evals Quick Start

• Google Sheets API Python Quickstart

• gspread Authentication Docs