Docs/consumer / integration openclaw

OpenClaw Integration Guide

OpenClaw is a modular framework for building "Skills-based" AI Agents. PipeAgent is a first-class citizen in the OpenClaw ecosystem, providing easy-to-use skill definitions.

Two ways to get feed skills

	PipeAgent Pilot (conversation)	CLI install (command line)
Best for	You already run OpenClaw and want the agent to search, compare metadata, and run install for you	Scripts, CI, or you prefer copy-paste `curl \	bash` from the dashboard
Discovery	Agent calls `/v1/catalog` and `/v1/search` with your read key	You pick a feed on the site or from API yourself
Metadata	Agent calls `/v1/feed/{id}/metadata` before installing	Optional; you can call the same API or read the feed page
Install	Agent runs `install.sh` in the shell (no read key in that command)	You run the same **`curl … \	bash`** one-liner (see CLI guide)
Read key	Required for JSON discovery/metadata APIs; not passed to `install.sh`	Same: key only for runtime / APIs, not for install

Shared requirements

PROVIDER_READ_API_KEY in the OpenClaw runtime (e.g. .env) so installed feed skills can call GET /v1/feed/{id}. Newer install.sh can prompt to write this automatically (default env path: ~/.openclaw/.env, overridable via --env-file). Where OpenClaw loads env vars: OpenClaw environment variables.

Default skill path: ~/.openclaw/workspace/skills/<folder>/SKILL.md — workspace ~/.openclaw/workspace must exist (or use --dest on install). Details: CLI install.

Next steps

PipeAgent Pilot — one-line curl … \| install-pipeagent-pilot.sh (or raw SKILL.md), then use chat to manage feeds.

CLI install — one-line install.sh, flags, Workbench copy block.

1. The SKILL.md standard

OpenClaw uses a directory-based skill structure. Each skill lives in its own folder and contains a SKILL.md file that acts as the agent's instruction manual.

text

~/.openclaw/workspace/
└── skills/
    └── your-skill-folder/
        └── SKILL.md

2. Manual per-feed template

If you want to customize a data skill by hand (not the pilot), use this shape:

yaml

---
name: "pa_custom_feed"
description: "PipeAgent: [Describe your data here]"
requires:
  env:
    - PROVIDER_READ_API_KEY
  config:
    - feed_id: "[Your-Feed-ID]"
---

### Instruction
You are an expert data analysis agent. Use this skill to fetch live structured data from PipeAgent.

1. Call the API using the configured feed_id.
2. Filter the JSON response to answer the user's specific question.
3. If multiple items are returned, summarize them in a bulleted list.

Generated skills from the feed Workbench include tool definitions and schema; see any feed’s OpenClaw tab.

3. Environment variables

bash

export PROVIDER_READ_API_KEY="your_api_key_from_dashboard"

OpenClaw merges env from the process, project .env, ~/.openclaw/.env, and config — see OpenClaw environment variables.

If you use install.sh, the script can now guide you to set PROVIDER_READ_API_KEY directly in OpenClaw env (default ~/.openclaw/.env), then prompt you to restart the gateway.

4. Why OpenClaw + PipeAgent?

Modular design: Swap data sources without changing your agent's core code.

Natural language discovery: OpenClaw agents can discover skills from YAML descriptions.

Instructional control: Use the Markdown section of SKILL.md to teach the agent how to interpret data.

Version 1.0.4 - Premium Infrastructure