POST
https://api.klarix.ai/v1/prospects/matchProspect Universe Match
Semantic vector search across 150,443 indexed B2B companies. Describe an ICP in prose and get ranked lookalikes with verified domains.
When to call it: You know the buyer you want and need the account list. Chain each hit into prospect-fit.
sub-100ms vector search · 200 to 600ms end to end including query embeddingMCP tool
find_matched_prospectsReturns
- Ranked company list with verified domains, verticals, and cosine similarity
- What they do, business model, target buyer, confirmed tech stack
- Dated growth and market signals
- A `suggested_action` naming the next call to chain into
Body parameters
| Name | Type | Req | Description |
|---|---|---|---|
| icp_or_query | string | required | Min 4 characters. ICP description, criteria, tech stack, or problem statement. Prose beats keywords: name the buyer and the problem, not a product category. |
| vertical | string | optional | Substring filter on the indexed vertical. Accepts any of the six labels listed under Index coverage, verbatim, e.g. "AI SaaS & DevTools". Filtering pre-empts the HNSW index, so expect a few hundred extra milliseconds. |
| limit | number | optional | Integer 1 to 25. Default 10. |
| min_similarity | number | optional | Cosine similarity floor, 0 to 1. Default 0.35. Raise to 0.5 for tight lookalikes; empty results usually mean this is too high. |
| exclude_domains | string[] | optional | Domains to drop from results — existing customers, disqualified accounts, yourself. |
| sub_category | string | optional | Narrower slice inside the vertical, e.g. "Serverless GPU Cloud", "Executive Search", "Contract Manufacturing & EMS". |
| headcount_min | number | optional | Positive integer. Drops accounts whose headcount band tops out below this. |
| headcount_max | number | optional | Positive integer. Drops accounts whose headcount band starts above this. |
| revenue_min | number | optional | Annual revenue floor in USD. Advisory only — the semantic index carries no revenue field, so it is recorded for downstream scoring rather than narrowing retrieval. |
| revenue_max | number | optional | Annual revenue ceiling in USD. Advisory only, same as `revenue_min`. |
| countries | string[] | optional | Up to 10. Restricts to accounts headquartered in these countries, matched against HQ location, e.g. ["United States", "Canada"]. |
| required_tech | string[] | optional | Up to 10. Keeps only accounts whose confirmed tech stack carries at least one of these, e.g. ["CUDA", "PyTorch"]. |
| our_company_name | string | optional | Your company. Talk tracks, landmines, and exploit vectors get written for your seller instead of a generic one. |
| our_domain | string | optional | Your domain, e.g. "klarix.ai". Also the label when no name is passed. |
| our_differentiators | string[] | optional | Up to 12 claims you win on. Each one gets tested against the evidence, and replaces any list loaded from `client_id`. |
| client_id | string | optional | Kebab-case Klarix client id. Loads your saved name, domain, differentiators, win themes, and ICP. Explicit fields always win. Unknown ids degrade the answer, never error. |
shell
curl -X POST https://api.klarix.ai/v1/prospects/match \
-H "Authorization: Bearer $KLARIX_API_KEY" \
-H "Content-Type: application/json" \
--max-time 240 \
-d '{
"icp_or_query": "B2B SaaS and technology companies with 50 to 500 employees scaling sales teams that need continuous competitive intelligence and prospect lookalikes without per-seat licensing",
"vertical": "AI SaaS & DevTools",
"limit": 3,
"min_similarity": 0.4,
"headcount_min": 50,
"headcount_max": 500,
"countries": [
"United States"
],
"required_tech": [
"Python"
],
"our_company_name": "Klarix",
"our_domain": "klarix.ai"
}'Request body (JSON)
{
"icp_or_query": "B2B SaaS and technology companies with 50 to 500 employees scaling sales teams that need continuous competitive intelligence and prospect lookalikes without per-seat licensing",
"vertical": "AI SaaS & DevTools",
"limit": 3,
"min_similarity": 0.4,
"headcount_min": 50,
"headcount_max": 500,
"countries": [
"United States"
],
"required_tech": [
"Python"
],
"our_company_name": "Klarix",
"our_domain": "klarix.ai"
}200 response (JSON envelope)
{
"success": true,
"data": {
"query": "B2B SaaS and technology companies with 50 to 500 employees scaling sales teams that need continuous competitive intelligence and prospect lookalikes without per-seat licensing",
"vertical": "AI SaaS & DevTools",
"total_matches": 3,
"total_universe_indexed": 150443,
"search_latency_ms": 74,
"source": "neon_pgvector",
"prospects": [
{
"company_name": "Modal",
"domain": "modal.com",
"vertical": "AI SaaS & DevTools",
"sub_category": "Serverless GPU Cloud",
"what_they_do": "Serverless cloud platform for running AI/ML models, batch inference, and fine-tuning with instant scaling.",
"target_buyer": "Head of AI, Staff ML Engineer, VP Engineering",
"business_model": "Usage-based compute & enterprise tiers",
"tech_stack": [
"Python",
"Rust",
"CUDA"
],
"growth_signals": [
"Hiring technical and sales roles",
"Enterprise tier launch in 2026"
],
"employee_count": "50-100",
"similarity": 0.921,
"relevance_score": 9,
"suggested_action": "POST /v1/intelligence/prospect-fit with domain \"modal.com\""
}
]
},
"markdown": "# Prospect Universe — Matched Accounts (5 found)\n\n**Query:** \"AI infrastructure companies with Python/Rust stacks\" | **Universe:** 150,443 accounts indexed | **Latency:** 75ms\n\n| Company & Domain | Vertical | Fit % | What they do | Key Tech |\n| --- | --- | --- | --- | --- |\n| 1. **Modal** (`modal.com`) | AI SaaS & DevTools | 92% | Serverless cloud platform for running AI/ML models | Python, Rust, CUDA |\n| 2. **Baseten** (`baseten.co`) | AI SaaS & DevTools | 88% | High-performance model inference infrastructure | Truss, Python, TensorRT |\n| 3. **Together AI** (`together.ai`) | AI SaaS & DevTools | 85% | Cloud platform for building and running open models | PyTorch, vLLM, Triton |\n\n**Recommended next moves:**\n- `score_prospect_fit(company=\"Modal\", domain=\"modal.com\")`\n- `get_competitor_battlecard(competitor_domain=\"modal.com\")`",
"meta": {
"latency_ms": 418,
"timestamp": "2026-09-02T14:22:11.402Z"
}
}Sample Output ReportMarkdown & Tables
Prospect Universe — Matched Accounts (5 found)
Query: "AI infrastructure companies with Python/Rust stacks" | Universe: 150,443 accounts indexed | Latency: 75ms
| Company & Domain | Vertical | Fit % | What they do | Key Tech |
|---|---|---|---|---|
1. Modal (modal.com) | AI SaaS & DevTools | 92% | Serverless cloud platform for running AI/ML models | Python, Rust, CUDA |
2. Baseten (baseten.co) | AI SaaS & DevTools | 88% | High-performance model inference infrastructure | Truss, Python, TensorRT |
3. Together AI (together.ai) | AI SaaS & DevTools | 85% | Cloud platform for building and running open models | PyTorch, vLLM, Triton |
Recommended next moves:
score_prospect_fit(company="Modal", domain="modal.com")get_competitor_battlecard(competitor_domain="modal.com")
