Building data-driven agents with OpenClaw and PipeAgent

Fuel AI agents with stable JSON APIs, not one-off scrapers: PipeAgent feeds install as OpenClaw Skills so your agent calls documented endpoints instead of parsing HTML.

Fast path: on any feed’s detail page, open the OpenClaw tab and copy the one-line install (installs into ~/.openclaw/workspace/skills/... by default; that workspace directory must exist). The script is hosted at https://pipeagent.dev/install.sh and talks to https://api.pipeagent.dev by default (override with PIPEAGENT_SITE for local dev—see comments in the script). Example:

curl -fsSL https://pipeagent.dev/install.sh | bash -s -- --alias "your-feed-endpoint-path"

You can also pass --id with the feed UUID, and optional --skillname, --dest, --key, or --env-file—see OpenClaw · CLI install.

Configure PROVIDER_READ_API_KEY in your OpenClaw runtime before the agent calls the feed API (OpenClaw environment variables). See OpenClaw Integration (Pilot vs CLI), PipeAgent Pilot, and CLI install for flags and URLs. The steps below are the longer manual path.

In this tutorial, we’ll build a simple "Market Intelligence Bot" that tracks the top 100 electronics products on Amazon and summarizes market sentiment.

Prerequisites

1. OpenClaw CLI installed (or your favorite IDE).

2. A PipeAgent read API key (Dashboard → keys).

3. Node.js or Python environment (depending on your OpenClaw setup).

Step 1: Initialize Your Agent

If you haven't already, create a new directory for your agent project:

mkdir my-data-agent
cd my-data-agent
mkdir skills

Step 2: Choose Your Feed

Open the Feeds directory and pick a feed. For this tutorial, we’ll use the Amazon Bestsellers - Electronics feed (or any feed with an OpenClaw tab).

On the Feed Detail page, click the "OpenClaw" tab. You'll see a snippet that looks like this:

---
name: "amazon_electronics_feed"
description: "PipeAgent: Fetches real-time structured data for top-performing electronics on Amazon."
requires:
  env:
    - PROVIDER_READ_API_KEY
  config:
    - feed_id: "amazon-electronics"
---

Step 3: Create the Skill

The fastest way to get your skill running is to use the "Download" button (next to the Copy icon) in the OpenClaw tab.

1. Download the SKILL.md file from PipeAgent (feed detail → OpenClaw).

2. Create a folder in your skills/ directory (e.g., skills/pa-amazon-electronics/).

3. Move the downloaded SKILL.md into that folder.

Alternatively, you can manually create the file:

### Instruction
You are a Market Intelligence Expert. Your job is to fetch the latest product rankings and identify shifts in consumer sentiment.

1. Fetch data from the "amazon_electronics_feed" using the `feed_id`.
2. Analyze the `rating_arrays` and `price_history`.
3. Highlight any product with a rating below 4.0 that is still in the top 10.
4. Provide a 3-sentence summary of the current market trend.

Step 4: Configure Environment Variables

OpenClaw needs your key to call the PipeAgent API (https://api.pipeagent.dev). Add it to .env or export it:

export PROVIDER_READ_API_KEY="your_api_key_from_dashboard"

Step 5: Test the Agent

Run your OpenClaw agent and ask it a question about the electronics market.

User: "What's the general mood in the electronics market today? Are there any top-sellers with suspicious quality?"

Agent: "Based on the latest data from the Amazon Bestsellers feed, consumer sentiment for the top 10 items is high (4.6 avg). However, the 'SuperBass 2000' headphones are currently ranked #4 despite a 3.8 rating—this suggests aggressive pricing is overriding quality concerns. Overall, the market remains dominated by Apple and Sony peripherals."

Why This Workflow Wins

Next Steps

Browse Feeds for crypto, e-commerce, GitHub trends, and more—each is a ready-made API for your next OpenClaw agent.

---

*For OpenClaw setup and orchestration, see the OpenClaw documentation.*