The server has 11 tools with mixed quality. Most tools lack complete parameter descriptions and schemas. Several tools are marked deprecated, signaling API churn. The 2 active search tools (web_search_exa, web_search_advanced_exa) have minimal descriptions (under 100 chars) and incomplete schemas. The server exposes the Exa API surface directly rather than composing simple, focused operations. Parameter schemas exist but lack type constraints and ranges. Error handling is not evident in the tool definitions. Tool composition is poor, deprecated tools (company_research_exa, people_search_exa, linkedin_search_exa, get_code_context_exa, deep_search_exa) clutter the interface, creating decision burden on LLMs. The deep_researcher_* tools are marked deprecated but still registered. Overall structure suggests early-stage implementation before rigorous tool design review.
[Deprecated: Use web_search_advanced_exa instead] Research any company to get business information, news, and insights. Best for: Learning about a company's products, services, recent news, or industry position. Returns: Company information from trusted business sources.
[Deprecated: Use web_fetch_exa instead] Extract content from specific URLs
[Deprecated] Check status and get results from a deep research task. Best for: Getting the research report after calling deep_researcher_start. Returns: Research report when complete, or status update if still running. Important: Keep calling with the same research ID until status is 'completed'.
[Deprecated] Start an AI research agent that searches, reads, and writes a detailed report. Takes 15 seconds to 2 minutes. Best for: Complex research questions needing deep analysis and synthesis. Returns: Research ID - use deep_researcher_check to get results. Important: Call deep_researcher_check with the returned research ID to get the report.
[Deprecated: Use web_search_advanced_exa instead] Deep search with query expansion and synthesized answers (requires API key)
5 of 11 tools explicitly marked [Deprecated] yet remain registered and discoverable. This creates confusion for LLMs about which tools to use and violates the single-responsibility principle. Deprecated tools should be removed from the schema entirely or moved to a legacy namespace.
Input schemas visible but lack type constraints and ranges. E.g., web_search_exa accepts 'numResults' as a number with no min/max bounds, LLMs can pass 0, 1000000, or negative values. web_search_advanced_exa schema not shown in source; cannot verify completeness.
Tool descriptions are too brief and lack context for tool selection. 'Real-time web search using Exa AI' (46 chars) doesn't explain when to use web_search_exa vs web_search_advanced_exa. Most descriptions under 100 chars, well below the 194-char baseline for A-grade tools. No descriptions explain dependencies, prerequisites, or return structure.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2025-06-18+ | v2 |
[Deprecated: Use web_search_exa instead] Search for code snippets, examples, and documentation from open source repositories
[Deprecated: Use web_search_advanced_exa instead]
[Deprecated: Use web_search_advanced_exa instead] Search for people and professional profiles
Extract content from specific URLs
Advanced web search with full Exa API control including category filters, domain restrictions, date ranges, highlights, summaries, and subpage crawling
Real-time web search using Exa AI
No documented output schemas for any tool. LLMs cannot infer what fields to expect or how to chain results. E.g., web_search_exa returns results but the response format is not documented, does it include URLs, snippets, metadata? downstream tools cannot know what to extract.
Parameter descriptions incomplete or missing. web_search_exa's 'query' param has description 'Search query' (13 chars), too vague. No guidance on query syntax, length limits, or special characters. Baseline is 72 chars per param description; most here are 20 - 50 chars.
Naming ambiguity: web_search_exa vs web_search_advanced_exa distinction unclear from names alone. Both are verb_noun, but the difference (basic vs advanced, filtering, summaries) is not evident in the names. Add qualifiers or split by use case to disambiguate.
deep_researcher_start and deep_researcher_check form a multi-step async pattern without documented error recovery. No guidance on timeout, retry logic, or what to do if research fails. The 'outputSchema' parameter is optional but not explained, LLMs may omit it when they should use it.
No evidence of error handling or recovery guidance in tool definitions. If a search returns no results, or fetch fails, the tool definitions don't specify what to return or how the LLM should respond. No guidance like 'Try broader search terms' or 'Check URL validity.'