KB Assistant · AI

Answers your customers will trust.

Voyage embeddings + reranking + cited sources. Built on the AI SDK with per-stage cost transparency. Every answer shows its work.

Try the assistant free See pricing
KB Assistant
How do I authenticate with the API?
You can authenticate using either OAuth tokens or personal API keys. Most production integrations use OAuth — see the auth guide for the full setup.
auth-guide.md api-tokens.md
Embed · $0.0001
Rerank · $0.0003
Answer · $0.0024
Total · $0.0028

Voyage embeddings

Better embeddings. Better recall.

Articles get chunked semantically and embedded with Voyage's voyage-3 model — purpose-built for retrieval, with measurably better recall than the OpenAI defaults that most docs platforms use. Your customers find the right chunk on the first try.

  • → Semantic chunking — paragraphs and code stay intact
  • → voyage-3 retrieval embeddings (1024-dim)
  • → Re-indexed automatically on every Git push
voyage-3 · 1024-dim

Reranker

From 20 candidates to the 5 that matter.

Vector retrieval is fast but coarse. We pull the top 20 candidates from your KB, then run a dedicated reranker over them to pick the 5 chunks that actually answer the question. The LLM only sees the best context, not the noise.

The reranker adds about 80ms but cuts hallucinations dramatically — and the cost is shown to you per query.

Top 20 candidates from vector search

Authentication: API keys 95
OAuth callback flow 92
Token rotation policy 89
Rate limiting headers 86
Webhook auth signatures 84
Public API surface 71
Refunding a charge 64
Webhook event types 58
Idempotency keys 52
Test mode 48

Cited answers, transparent costs

Every claim has a source. Every query has a price tag.

Answers cite the chunks they came from. Click a citation to jump straight to the source article. And every query exposes its cost broken down by stage — embed, rerank, answer — so you can budget exactly.

Most platforms hide the cost behind a flat "AI add-on" fee. We show you the unit economics, charge fairly, and let you set per-org caps so a viral thread can't blow up your bill.

Admin · KB Assistant

Per-stage cost · last 7 days

Embed
$1.42
Rerank
$2.04
Answer
$5.84
Total $9.30

Across 1,142 queries · avg $0.0081/query

Compare the pipeline

Mintlify, ReadMe, GitBook all do basic RAG.

KBKit goes further on the bits that actually affect answer quality and bill predictability.

Capability KBKit Mintlify ReadMe GitBook
Cited sources on every answer ✓ ✓ ~ ~
Per-stage cost transparency ✓ ✕ ✕ ✕
Reranker (top-N → top-K) ✓ ✕ ✕ ✕
Streaming responses ✓ ✓ ✓ ✓
Per-org cost caps ✓ ✕ ✕ ✕

~ partial · based on publicly available product documentation as of Apr 2026.

Ready to ship answers your customers can trust?

Free during early access. No credit card required.

Try it free See pricing