exa-search ยท diff
git:20260307.9e17ff6 to git:20260409.c75add8
32 added, 50 removed. Audit A to A.
---
name: exa-search
- description: Search the web, code, companies, and people using Exa's AI-powered search engine. Use this skill whenever the user asks to search for current information, find code examples, research a company, find professional profiles, or perform deep research on a complex topic. This skill is superior to basic web search as it provides clean, LLM-ready content and specialized search modes.
+ description: Search the web, read webpages as markdown, and run filtered web retrieval with Exa MCP. Use this skill whenever the user asks for current web information, web research, domain/date/category-filtered search, company or people lookup via search filters, or extracting clean page content from one or more URLs.
---
- # Exa Search Skill
+ # Exa Search
- This skill provides access to Exa's AI-powered search capabilities via their hosted MCP server.
+ Use this skill for Exa MCP web retrieval.
## When to Use
- - **Web Search**: For current events, facts, or general knowledge.
- - **Code Context**: To find documentation, API usage, or code examples from GitHub, Stack Overflow, etc.
- - **Company Research**: To get business insights, news, and metadata about any company.
- - **People Search**: To find professional profiles and contact info.
- - **Deep Research**: For complex questions that require multiple searches and synthesis (takes 15s - 3min).
- - **Crawling**: To extract full text content from a specific known URL.
-
- ## Available Tools
-
- The skill provides the following tools via the `scripts/query.py` script:
-
- - `web_search_exa`: General web search.
- - `web_search_advanced_exa`: Search with filters (domain, date, category).
- - `get_code_context_exa`: Targeted programming and documentation search.
- - `company_research_exa`: Research companies.
- - `people_search_exa`: Find professional profiles.
- - `crawling_exa`: Extract content from a URL.
- - `deep_researcher_start`: Start a long-running research task.
- - `deep_researcher_check`: Get results from a research task.
+ - Current web search
+ - Web research that needs clean LLM-ready results
+ - Domain, date, and category filtered search
+ - Company or people lookup through search filters
+ - Read full page content from known URLs as markdown
- ## Usage Guide
+ ## Preferred Tools
+ - `web_search_exa`: default search tool
+ - `web_fetch_exa`: fetch one or more URLs as markdown
+ - `web_search_advanced_exa`: use only when filters or advanced options are needed
- ### Calling a Tool
- Use the `scripts/query.py` script to interact with Exa:
+ ## Workflow
+ 1. Run `list_tools` to confirm what the current server exposes.
+ 2. For most tasks, start with `web_search_exa`.
+ 3. If the user already has URLs, or search results need full content, use `web_fetch_exa`.
+ 4. Use `web_search_advanced_exa` only when the task needs filters such as domains, categories, dates, highlights, summaries, or subpage crawling.
+ 5. Prefer the current tools above as the main operating path.
+ ## Commands
```bash
- # List all available tools and their schemas
+ # Inspect current MCP tools
python3 /var/minis/skills/exa-search/scripts/query.py list_tools
- # Search the web
- python3 /var/minis/skills/exa-search/scripts/query.py call_tool web_search_exa '{"query": "latest news about SpaceX Starship"}'
+ # General search
+ python3 /var/minis/skills/exa-search/scripts/query.py call_tool web_search_exa '{"query":"latest changes in Exa MCP docs","numResults":5}'
- # Search for code examples
- python3 /var/minis/skills/exa-search/scripts/query.py call_tool get_code_context_exa '{"query": "React server components example"}'
+ # Fetch page content
+ python3 /var/minis/skills/exa-search/scripts/query.py call_tool web_fetch_exa '{"urls":["https://exa.ai/docs/reference/exa-mcp"]}'
- # Perform deep research
- # 1. Start the research
- python3 /var/minis/skills/exa-search/scripts/query.py call_tool deep_researcher_start '{"instructions": "Research the impact of Llama 3 on the open source AI ecosystem"}'
- # 2. Get the researchId from the output, then poll for results
- python3 /var/minis/skills/exa-search/scripts/query.py call_tool deep_researcher_check '{"researchId": "..."}'
+ # Advanced search with filters
+ python3 /var/minis/skills/exa-search/scripts/query.py call_tool web_search_advanced_exa '{"query":"Exa MCP documentation","includeDomains":["exa.ai"],"numResults":5,"enableHighlights":true}'
```
- ### Parameters Reference
-
- #### web_search_exa
- - `query` (string, required): The search query.
- - `numResults` (number): Default 8.
- - `livecrawl` (string): 'fallback' (default) or 'preferred'.
-
- #### get_code_context_exa
- - `query` (string, required): Programming question or API name.
- - `tokensNum` (number): 1000-50000, default 5000.
-
- #### deep_researcher_start
- - `instructions` (string, required): Detailed research question.
- - `model` (string): 'exa-research-fast' (default), 'exa-research', or 'exa-research-pro'.
+ ## Notes
+ - Base MCP URL: `https://mcp.exa.ai/mcp`
+ - Do not assume optional tools are enabled in every deployment; verify with `list_tools`.
+ - If the server exposes only the default tools, use `web_search_exa` + `web_fetch_exa` as the primary path.
## Environment Variables
- - `EXA_API_KEY`: Required. If not set, prompt the user to [Set EXA_API_KEY](minis://settings/environments?create_key=EXA_API_KEY&create_value=).
+ - `EXA_API_KEY`: optional for basic use, recommended for higher limits and better reliability.
+ - If missing when needed, prompt the user to [Set EXA_API_KEY](minis://settings/environments?create_key=EXA_API_KEY&create_value=).