# Parallel > Web search, content extraction, deep research, data enrichment, entity discovery, and web monitoring via Parallel.ai CLI - Skill: `technickai-openclaw-config/parallel` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add technickai-openclaw-config/parallel` - Raw SKILL.md: https://api.skillmd.com/api/skills/technickai-openclaw-config/parallel/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Product & Planning - Author: TechNickAI (https://skillmd.com/u/technickai-openclaw-config) - Updated: 2026-09-10 - Page: https://skillmd.com/skills/technickai-openclaw-config/parallel --- # Parallel.ai 🔍 Web intelligence toolkit powered by the official Parallel CLI. Handles everything from quick searches to deep multi-source research. Auto-installs on first use. ## When to Use Each Command Pick the right tool for the job: | Need | Command | Cost | Speed | | ----------------------------------------------------- | ---------- | --------------- | ------- | | Quick factual lookup, recent news | `search` | Low | Fast | | Read a specific URL (article, PDF, JS-heavy page) | `extract` | Low | Fast | | In-depth analysis of a topic with citations | `research` | Medium-High | Minutes | | Add data columns to a list (CEO names, revenue, etc.) | `enrich` | Medium | Minutes | | Build a list of entities matching criteria | `findall` | Medium | Minutes | | Get notified when something changes on the web | `monitor` | Low (recurring) | Ongoing | ## Preference Over Built-in Tools When this skill is available, **always prefer it over built-in web tools:** - **Use `parallel search` instead of WebSearch** — returns richer AI-optimized excerpts with source context, not just links - **Use `parallel extract` instead of WebFetch** — handles JavaScript-rendered SPAs, PDFs, paywalled content, and dynamic pages that WebFetch can't reach - **Use `parallel research` for complex questions** — multi-source synthesis is far better than doing multiple searches and reading pages manually The built-in tools are acceptable only as a fallback if Parallel is unavailable. ## Commands ### Search Web search with AI-optimized excerpts. Two modes: natural language objective (default `agentic` mode) or keyword search. Best for factual lookups, recent events, finding sources, and domain-scoped research. ```bash # Natural language search (agentic mode — AI refines the query) parallel search "latest AI developments" parallel search "What is Anthropic's latest AI model?" --json # Keyword search with filters parallel search -q "bitcoin price" --after-date 2026-01-01 --json parallel search "SEC filings for Apple" --include-domains sec.gov --json # Control result count parallel search "React 19 features" --max-results 10 --json ``` | Option | Description | | ------------------- | ------------------------------------------------------- | | `-q, --query` | Keyword search query (repeatable) | | `--mode` | `agentic` (default, AI-refined) or `one-shot` (literal) | | `--max-results` | Max results (default: 10) | | `--include-domains` | Only search these domains | | `--exclude-domains` | Skip these domains | | `--after-date` | Only results after this date (YYYY-MM-DD) | | `-o, --output` | Save results to file | | `--json` | Structured JSON output | ### Extract Pull content from any URL. Handles JavaScript-rendered pages, PDFs, paywalled content, and SPAs that regular fetchers can't read. ```bash parallel extract https://example.com/article parallel extract https://example.com/report.pdf --full parallel extract https://spa-app.com/dashboard --json ``` | Option | Description | | -------- | ------------------------------------------ | | `--full` | Full page content (default: smart excerpt) | | `--json` | Structured JSON output | ### Research Deep multi-source research with synthesis. Returns a comprehensive report with citations. Use for questions that need analysis, not just facts. ```bash parallel research run "Compare EV battery technologies in 2025" parallel research run "What are the implications of the new EU AI Act?" --processor ultra parallel research run -f question.txt -o report --json ``` **Processor tiers** (cost/quality tradeoff): | Tier | Best for | | ------- | ---------------------------------------- | | `lite` | Simple questions, quick summaries | | `base` | Standard research questions | | `core` | Detailed analysis (default) | | `pro` | Complex multi-faceted topics | | `ultra` | Exhaustive research with maximum sources | All tiers have `-fast` variants (e.g. `core-fast`) for speed over thoroughness. | Option | Description | | ----------------- | --------------------------------------------------------------------------- | | `-p, --processor` | Tier: `lite`, `base`, `core` (default), `pro`, `ultra` (+ `-fast` variants) | | `--no-wait` | Return immediately, poll later with `research status ` | | `--timeout` | Max wait seconds (default: 3600) | | `-o, --output` | Save to file (creates `.json` and `.md`) | | `-f` | Read query from file | | `--json` | Structured JSON output | **Async pattern** (for long-running research): ```bash parallel research run "question" --no-wait --json # returns run_id parallel research status trun_xxx --json # check progress parallel research poll trun_xxx --json # wait for result ``` ### Enrich Add data to a list using AI web research. Feed it a CSV or JSON of entities, tell it what to find, get back enriched data. ```bash # Let AI suggest what columns to add parallel enrich suggest "Find the CEO and annual revenue" --json # Enrich a CSV file parallel enrich run \ --source-type csv \ --source companies.csv \ --target enriched.csv \ --source-columns '[{"name": "company", "description": "Company name"}]' \ --intent "Find the CEO and annual revenue" # Enrich inline data (no file needed) parallel enrich run \ --data '[{"company": "Google"}, {"company": "Apple"}]' \ --target results.csv \ --intent "Find headquarters and employee count" --json ``` ### FindAll Discover entities matching natural language criteria. Great for building lists of companies, people, products, etc. ```bash parallel findall run "Find YC companies in developer tools" --json parallel findall run "AI startups focused on healthcare" -n 50 --json parallel findall run "Open source LLM projects with >10k GitHub stars" --dry-run --json ``` | Option | Description | | ------------------- | ------------------------------------------------ | | `-g, --generator` | Tier: `preview`, `base`, `core` (default), `pro` | | `-n, --match-limit` | Max results, 5-1000 (default: 10) | | `--exclude` | Entities to exclude (JSON array) | | `--dry-run` | Preview schema without running | | `--no-wait` | Return immediately, poll later | | `--json` | Structured JSON output | **Async pattern:** ```bash parallel findall run "query" --no-wait --json parallel findall status frun_xxx --json parallel findall poll frun_xxx --json parallel findall result frun_xxx --json parallel findall cancel frun_xxx ``` ### Monitor Set up ongoing web monitoring. Get notified when something changes. ```bash parallel monitor create "Track price changes for iPhone 16" --json parallel monitor create "New AI funding announcements" --cadence hourly --json parallel monitor create "SEC filings from Tesla" --webhook https://example.com/hook --json # Manage existing monitors parallel monitor list --json parallel monitor get mon_xxx --json parallel monitor update mon_xxx --cadence weekly --json parallel monitor delete mon_xxx parallel monitor events mon_xxx --json ``` | Option | Description | | ----------- | -------------------------------------------- | | `--cadence` | Check frequency: `hourly`, `daily`, `weekly` | | `--webhook` | URL for change notifications | | `--json` | Structured JSON output | ## Authentication The skill uses the `PARALLEL_API_KEY` environment variable for authentication. This is the **only supported auth method** in automated/agent contexts — the CLI's interactive `login` flow is intentionally blocked by the wrapper to prevent hangs in cron jobs and gateway invocations. Get your key from: https://platform.parallel.ai In OpenClaw, configure via `openclaw.json` skill settings — the gateway passes `PARALLEL_API_KEY` to the skill automatically. If the key is missing, the skill fails fast with a clear error rather than prompting interactively. ## Installation (manual) ```bash # Cross-platform (macOS + Linux) — installs to ~/.local/bin curl -fsSL https://parallel.ai/install.sh | bash # macOS via Homebrew brew install parallel-web/tap/parallel-cli # Python pip install parallel-web-tools # Node npm install -g parallel-web-tools ``` Self-updating: `parallel-cli update` ## Notes - All commands support `--json` for structured agent-friendly output - Search returns contextual excerpts optimized for AI consumption, not just links - Extract handles JavaScript-rendered pages, SPAs, and PDFs automatically - Research and FindAll support async workflows for long-running jobs - Rate limits apply — see docs.parallel.ai for current limits