High-accuracy web research platform via Parallel.ai, built for AI agents. 7 APIs covering search, extraction, deep research, chat, entity discovery, monitoring, and batch execution.
- Search - high-accuracy web search with 3 modes: one-shot, fast (~1s), agentic (multi-hop)
- Extract - pull clean content from URLs, JS-rendered pages, and PDFs
- Task (Deep Research) - 8 processor tiers (lite through ultra8x), MCP tool calling, authenticated browsing
- Chat - OpenAI-compatible chat completions with web grounding and basis citations
- FindAll - entity discovery at web scale with 4 generator tiers (preview, base, core, pro)
- Monitor - scheduled web change tracking with webhook notifications
- Task Groups - batch up to 1,000 research tasks per POST with SSE streaming
git clone https://github.com/mvanhorn/clawdbot-skill-parallel.git ~/.openclaw/skills/parallel
cd ~/.openclaw/skills/parallel
pip install -r requirements.txtGet a key from Parallel.ai, then:
export PARALLEL_API_KEY="your-key-here"- "Use Parallel to research transformer architectures"
- "Deep search for the latest on AI regulation in the EU"
- "Find all AI startups that raised Series A in the last 6 months"
- "Monitor AI safety news daily and alert me"
- "Chat with web grounding about recent funding rounds"
- "Batch research the top 10 cloud providers"
| API | Endpoint | Description |
|---|---|---|
| Search | POST /v1/search |
Web search with one-shot, fast, agentic modes |
| Extract | POST /v1beta/extract |
Content extraction from URLs and PDFs |
| Task | POST /v1/tasks/runs |
Deep research with 8 processor tiers |
| Chat | POST /v1/chat/completions |
OpenAI-compatible with web grounding |
| FindAll | POST /v1beta/findall/runs |
Entity discovery at scale |
| Monitor | POST /v1alpha/monitors |
Scheduled web change tracking |
| Task Groups | POST /v1beta/tasks/groups |
Batch up to 1,000 tasks |
- Python:
pip install parallel-web(v0.4.2) - TypeScript:
npm install parallel-web - CLI:
brew install parallel-web/tap/parallel-cli - Vercel AI SDK:
npm install @parallel-web/ai-sdk-tools
- API key is loaded from the
PARALLEL_API_KEYenvironment variable only - never hardcoded - All user input is safely escaped before being sent to the API (no JSON injection)
- Dependencies are pinned in
requirements.txt - When using authenticated sources (BROWSERUSE_API_KEY), be aware that your key is transmitted to Parallel.ai's servers which proxy it to browser-use.com
- Docs: https://docs.parallel.ai
- Platform: https://platform.parallel.ai
MIT