gtm-now 0.10.5 → 0.10.6

package/dist/index.js

@@ -10784,7 +10784,7 @@ var init_prospect = __esm({
     prospectTools = [
       {
         name: "gtm_find_companies",
-        description: "Discover companies matching ICP criteria via DiscoLike, Apollo, or StoreLeads. Supports lookalike discovery, ICP text, tech stack filtering. Use gtm_tool_help for full parameter details.",
+        description: 'Discover companies matching ICP criteria via DiscoLike, Apollo, StoreLeads, or Exa (semantic/neural search). Use provider: "exa" for meaning-based search that understands intent, not just keywords.',
         inputSchema: {
           type: "object",
           properties: {
@@ -11526,7 +11526,15 @@ For DTC/e-commerce brands, use provider: "storeleads" \u2014 it has revenue esti
 
 Tip: DiscoLike lookalike mode is extremely powerful. If you have 2-3 example companies, use the domains parameter to find hundreds of similar ones.
 
-Single domain lookup: passing just domain (no icp_text, domains, category, tech_stack, or country) does a direct company profile fetch with rich data (growth signals, engagement metrics, tech stack from vendors). This is for researching ONE specific company, not discovery.`,
+Single domain lookup: passing just domain (no icp_text, domains, category, tech_stack, or country) does a direct company profile fetch with rich data (growth signals, engagement metrics, tech stack from vendors). This is for researching ONE specific company, not discovery.
+
+Exa (provider: "exa") \u2014 semantic/neural search:
+- Meaning-based ICP matching: "B2B SaaS companies selling to healthcare, not healthtech vendors" actually works
+- Best for: nuanced ICP descriptions where keyword matching fails
+- Uses icp_text for semantic search, domains for lookalike discovery
+- Does NOT return employee count, industry, or tech stack \u2014 use for discovery, then enrich with DiscoLike or Apollo
+- Must be explicitly selected with provider: "exa" (not auto-selected)
+- Requires GTM_EXA_API_KEY env var`,
       gtm_find_people: `Find contacts at companies by job title, company, or domain. Returns name, title, email (when available), phone, and LinkedIn URL.
 
 Powered by DiscoLike Contacts, Apollo, or BetterContact depending on configuration.