tinyfish-openclaw

TinyFish for OpenClaw

AI-powered web automation integrated with OpenClaw via MCP protocol.

Quick Start

The TinyFish MCP server is already configured and running. You can use it directly through OpenClaw commands.

Basic Usage

# Scrape a website
tinyfish scrape "https://example.com" "Extract product names and prices"

# Automate a task
tinyfish automate "https://example.com/contact" "Fill form with name John Doe and submit"

Available Commands

1. tinyfish scrape

Extract data from websites using natural language.

Syntax:

tinyfish scrape <url> <goal>

Examples:

# Extract product information
tinyfish scrape "https://store.com/products" "Extract first 10 product names, prices, and ratings as JSON"

# Collect contact information
tinyfish scrape "https://directory.com" "Extract company names, emails, and phone numbers"

# Aggregate news
tinyfish scrape "https://news-site.com" "Extract top 5 article titles, authors, and URLs"

# Research data
tinyfish scrape "https://scholar.google.com/scholar?q=AI" "Extract paper titles, authors, and citation counts"

# Price monitoring
tinyfish scrape "https://competitor.com/product" "Extract product name, current price, and discount"

2. tinyfish automate

Automate tasks on websites.

Syntax:

tinyfish automate <url> <goal>

Examples:

# Form filling
tinyfish automate "https://example.com/contact" "Fill name 'John Doe', email 'john@example.com', and submit"

# Login and extract
tinyfish automate "https://example.com/login" "Login with username 'test' password 'pass123', go to dashboard, extract balance"

# Multi-step workflow
tinyfish automate "https://example.com/signup" "Fill signup form, accept terms, click register"

# Search and extract
tinyfish automate "https://example.com" "Search for 'laptop', click search, extract first 5 results"

Integration with OpenClaw Workflows

Workflow 1: Daily Price Monitoring

Create a cron job in OpenClaw:

# .openclaw/workflows/price-monitor.yml
name: Daily Price Monitor
schedule: "0 9 * * *"  # Every day at 9 AM
steps:
  - name: Check Competitor Prices
    command: tinyfish scrape "https://competitor.com/products" "Extract all product names and prices as JSON"
  - name: Compare with Our Prices
    command: compare-prices --input $PREVIOUS_OUTPUT
  - name: Send Alert
    command: notify --channel telegram --message "Price changes detected"

Workflow 2: Lead Generation

# .openclaw/workflows/lead-gen.yml
name: Lead Generation
trigger: manual
steps:
  - name: Scrape Directory
    command: tinyfish scrape "https://business-directory.com" "Extract company names, emails, phones, websites"
  - name: Save to Database
    command: save-to-db --table leads --data $PREVIOUS_OUTPUT
  - name: Send to CRM
    command: crm-sync --source leads

Workflow 3: Content Aggregation

# .openclaw/workflows/content-aggregator.yml
name: Content Aggregator
schedule: "0 */6 * * *"  # Every 6 hours
steps:
  - name: Scrape News Site 1
    command: tinyfish scrape "https://news1.com" "Extract top 10 articles with titles, authors, URLs"
  - name: Scrape News Site 2
    command: tinyfish scrape "https://news2.com" "Extract top 10 articles with titles, authors, URLs"
  - name: Merge Results
    command: merge-json --files $STEP1_OUTPUT $STEP2_OUTPUT
  - name: Post to Channel
    command: post-to-discord --channel news --data $PREVIOUS_OUTPUT

Advanced Usage

Using with OpenClaw Scripts

#!/bin/bash
# monitor-prices.sh

# Scrape competitor prices
PRICES=$(tinyfish scrape "https://competitor.com/products" "Extract product names and prices as JSON")

# Parse and compare
echo "$PRICES" | jq '.items[] | select(.price < 100)'

# Send alert if needed
if [ $? -eq 0 ]; then
    openclaw notify --message "Low prices detected!"
fi

Using with OpenClaw Agents

# openclaw-agent.py
import subprocess
import json

def scrape_website(url, goal):
    result = subprocess.run(
        ['openclaw', 'tinyfish', 'scrape', url, goal],
        capture_output=True,
        text=True
    )
    return json.loads(result.stdout)

# Usage
data = scrape_website(
    'https://example.com',
    'Extract product information'
)

print(f"Found {len(data['items'])} products")

Configuration

MCP Server Settings

The TinyFish MCP server is configured in OpenClaw's config:

{
  "mcp": {
    "servers": {
      "tinyfish": {
        "command": "node",
        "args": ["E:\\ZAHRA-WORKSPACE\\mcp-servers\\tinyfish\\server.js"],
        "env": {
          "TINYFISH_API_KEY": "sk-tinyfish-F2ZFleAQ680jRwwSB2UKLuimrAJuELaT",
          "PORT": "8080"
        }
      }
    }
  }
}

Environment Variables

Set in .openclaw/env:

TINYFISH_API_KEY=sk-tinyfish-F2ZFleAQ680jRwwSB2UKLuimrAJuELaT
TINYFISH_URL=http://localhost:8080

Best Practices

1. Write Clear Goals

❌ Bad: "Get data" ✅ Good: "Extract product names, prices, and ratings as JSON array"

2. Specify Output Format

❌ Bad: "Extract products" ✅ Good: "Extract products as JSON with fields: name, price, inStock"

3. Limit Scope

❌ Bad: "Extract all products" ✅ Good: "Extract first 10 products"

4. Handle Errors

if tinyfish scrape "$URL" "$GOAL"; then
    echo "Success"
else
    echo "Failed, retrying..."
    sleep 5
    tinyfish scrape "$URL" "$GOAL"
fi

5. Use Caching

# Cache results for 1 hour
CACHE_FILE="/tmp/tinyfish-cache-$(echo $URL | md5sum | cut -d' ' -f1)"

if [ -f "$CACHE_FILE" ] && [ $(($(date +%s) - $(stat -c %Y "$CACHE_FILE"))) -lt 3600 ]; then
    cat "$CACHE_FILE"
else
    tinyfish scrape "$URL" "$GOAL" | tee "$CACHE_FILE"
fi

Common Use Cases

1. Price Monitoring Bot

# Check prices every hour
*/60 * * * * tinyfish scrape "https://competitor.com" "Extract prices" | openclaw notify

2. Lead Generation

# Generate leads from directory
tinyfish scrape "https://directory.com" "Extract contacts" | openclaw save-leads

3. Content Aggregation

# Aggregate news from multiple sources
for site in news1.com news2.com news3.com; do
    tinyfish scrape "https://$site" "Extract articles"
done | openclaw merge-and-post

4. Research Data Collection

# Collect academic papers
tinyfish scrape "https://scholar.google.com/scholar?q=$QUERY" "Extract papers" | openclaw save-research

5. Market Intelligence

# Monitor competitor activities
tinyfish scrape "https://competitor.com/blog" "Extract latest posts" | openclaw analyze-trends

Troubleshooting

Server Not Responding

# Check server status
curl http://localhost:8080/health

# Restart server
cd E:\ZAHRA-WORKSPACE\mcp-servers\tinyfish
npm start

Timeout Errors

• Increase timeout in OpenClaw config
• Simplify the goal
• Try again (some sites are slow)

Empty Results

• Make goal more specific
• Check if URL is correct
• Verify website is accessible

Performance

• Average execution time: 15-30 seconds
• Success rate: High (depends on website)
• Concurrent requests: Supported
• Rate limiting: Handled automatically

Resources

• TinyFish Docs: https://docs.tinyfish.ai/
• OpenClaw Docs: E:\ZAHRA-WORKSPACE\docs
• MCP Server: E:\ZAHRA-WORKSPACE\mcp-servers\tinyfish
• Support: https://discord.gg/tinyfish

License

MIT License

Author

Zahra AI - AI Chief Digital Officer
Created: 2026-03-29