[exa-mcp-server]: use highlights: true across all search tools

[devin-ai-integration]

Summary

Switches search tools from requesting full text/context content to highlights: true, which returns focused relevant snippets per search result.

Tools updated:

• web_search_exa: contents: { text: true, context: { maxCharacters: 10000 } } → contents: { highlights: true } (also removed contextMaxCharacters param)
• company_research_exa: contents: { text: { maxCharacters: 7000 } } → contents: { highlights: true }
• people_search_exa: contents: { text: { maxCharacters: ... } } → contents: { highlights: true }
• linkedin_search_exa: contents: { text: { maxCharacters: ... } } → contents: { highlights: true }

Unchanged:

• crawling_exa — the /contents endpoint does not support highlights: true (returns text regardless), so this tool remains on text: { maxCharacters } with sanitizeContentsResponse
• web_search_advanced_exa — excluded per request
• deep_search_exa — already uses highlights with its own config

Updated tools now return a formatted string with Title/URL/Published/Highlights per result (separated by ---) instead of raw JSON.stringify or the server-generated context blob.

Also updates ExaSearchRequest type in types.ts to replace context with highlights.

Updates since last revision

• Rebased onto main to resolve merge conflicts with the recently-added sanitizeSearchResponse/sanitizeContentsResponse utilities.
• Removed sanitizeSearchResponse imports from companyResearch, peopleSearch, and linkedInSearch since these tools now return formatted plaintext highlights instead of JSON.stringify(response.data).
• crawling_exa retains sanitizeContentsResponse since it still returns JSON.

Review & Testing Checklist for Human

• sanitizeSearchResponse is bypassed for updated tools — the recently-merged sanitizer is no longer called for companyResearch, peopleSearch, and linkedInSearch because they now format results as plaintext. Verify the sanitizer wasn't protecting against anything critical (e.g., stripping sensitive fields from the API response).
• Response format is a breaking change — companyResearch, peopleSearch, and linkedInSearch previously returned JSON.stringify(response.data). They now return formatted plaintext (Title:\nURL:\nHighlights:\n...). Any MCP clients parsing JSON from these tools will break.
• contextMaxCharacters param removed from web_search_exa — existing clients passing this param will get a schema validation error.
• Test with an actual MCP client (e.g. Claude Desktop) to verify the formatted highlights output is useful and sufficient for LLM consumption. Suggested: invoke each updated tool once and compare output quality to the previous version.

Notes

• ExaSearchRequest.contents.highlights type supports maxCharacters, numSentences, highlightsPerUrl, and query fields, though only true (boolean) is used by the updated tools.
• API testing confirmed highlights: true works correctly on /search but NOT on /contents (crawling endpoint).

Link to Devin session: https://app.devin.ai/sessions/828583886cd4479484cf6df638acdd7d

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
exa-mcp-server	Ready	Preview, Comment	Mar 19, 2026 8:13pm