Examples & References
Patterns and multi-step workflows. All calls use ~~capability with fallback (see CONNECTORS.md).
Reference Files
| File | Description |
|---|---|
| tool-patterns.md | Tool call patterns organized by capability |
| workflow-examples.md | Multi-step workflow examples |
| competitive-analysis.md | Scenario: competitor comparison |
| market-research.md | Scenario: market research |
| technical-audit.md | Scenario: technical audit |
| person-company-lookup.md | Scenario: person/company lookup |
| topic-deep-dive.md | Scenario: topic deep dive |
| news-trends.md | Scenario: news & trends |
Quick Reference: Capabilities
| Capability | What it does | Fallback order |
|---|---|---|
~~search |
Find pages on the web | web_search_exa → perplexity_search → search_web → firecrawl_search |
~~answer |
Short AI answer with citations | perplexity_ask → perplexity_reason → perplexity_search / ~~search results + your own cited synthesis |
~~scrape |
Read a single page (with question: only the relevant passages) |
read_url → firecrawl_scrape → web_fetch_exa (maxCharacters: 20000) |
~~batch_search |
Search multiple queries (≤5 per call) | search_web with an array → one web_search_exa per query |
~~batch_scrape |
Read multiple pages (≤5 per call) | read_url with an array → web_fetch_exa (maxCharacters: 20000) → one firecrawl_scrape per URL (fallbacks return full pages, no question) |
~~crawl |
Crawl entire site | firecrawl_crawl → firecrawl_map + batch scrape |
~~extract |
Structured data extraction | firecrawl_scrape (JSON format) → firecrawl_agent for unknown URLs |
~~academic_search |
Scientific papers | firecrawl_research_search_papers → search_arxiv / search_ssrn → perplexity_search |
~~code_search |
Code examples, issues, docs | firecrawl_developer_search → web_search_advanced_exa (opt-in) → search + "github" |
~~deep_agent |
Heavy research pass for depth deep |
agent_run → firecrawl_agent → perplexity_research |
Quick Workflow Patterns
Quick Search with Fallback
1. ~~search(query) → results
2. If error → next provider automatically
3. ~~scrape(best_url) → content
4. If error → next provider automatically
Parallel Research Batch
1. Plan related terms and 3-5 queries yourself
2. ~~batch_search(queries[]) → batch results
3. Rank by relevance → top results
4. ~~batch_scrape(top_5, question, topk) → only the relevant passages
5. Deduplicate → clean data
Full 7-Step Research
1. CLASSIFY → research type + depth
2. PLAN → 3-7 queries from different angles
3. SEARCH → ~~batch_search / ~~search (with fallback); depth deep adds ~~deep_agent
4. READ → ~~batch_scrape top-5 with a question (with fallback)
5. EXTRACT → key facts, data, quotes
6. SYNTHESIZE → deduplicate + cross-check
7. REPORT → template + methodology
Conventions
- All workflows use
~~capabilityplaceholders — see CONNECTORS.md for provider mapping - Fallback chains apply automatically on errors or empty results
- All reports include Methodology section
- Every fact must have a URL source