parallel

What This Skill Does

The Parallel skill integrates the Parallel.ai research engine directly into your OpenClaw agent. It provides high-accuracy web search and deep research capabilities, allowing your agent to move beyond basic LLM knowledge. By leveraging Parallel's specialized API, the agent can perform everything from rapid fact-checking and news retrieval to complex, multi-hop deep research and automated entity discovery. The skill is designed specifically for AI agents, offering clean content extraction, precise citations, and support for authenticated, private data sources.

Installation

To install this skill, use the ClawHub CLI inside your project environment: clawhub install openclaw/skills/skills/mvanhorn/parallel

Ensure you have the required dependencies installed via pip: pip install parallel-sdk requests

You must also obtain an API key from https://platform.parallel.ai and set it as an environment variable: export PARALLEL_API_KEY="your-key"

If you require research across authenticated or private web sources, ensure your BROWSERUSE_API_KEY is also configured.

Use Cases

This skill is perfect for research-heavy workflows. Use the Search API for time-sensitive lookups and current events. Use the Task API for deep enrichment, such as pulling historical financial data or generating comprehensive market analysis reports. The FindAll API acts as an automated lead-generation and data-gathering engine, transforming natural language requests like "find all dental practices in a specific region" into structured, cited datasets. Finally, the Extract API can be used to ingest clean content from complex URLs or PDF files for processing.

Example Prompts

1. "Perform a deep market research report on the growth of decentralized energy grids in Europe over the last three years, including citations for all claims."
2. "Find all Series A AI startups from 2025 that focus on healthcare, and extract their funding totals and current employee counts into a JSON format."
3. "Search for the latest updates on the NXP K66 microcontroller and extract the specific hardware performance specs from the official product page."

Tips & Limitations

• Processor Selection: Match your workload to the right engine. Use base for quick lookups to save costs and time, but switch to ultra for long-form, multi-hop research reports.
• Authentication: If you are scraping behind-login content, ensure you have the browser-use key configured to avoid 403 errors.
• Asynchronous Jobs: The FindAll API is non-blocking. For large datasets, use the --status command to track progress rather than expecting an immediate response.