Coinbase x402 · Vector DB · Gemini

The agentic memory
LLM gateway.

A high-performance LLM gateway with built-in memory designed for agents. Pay as you go. Remember everything. No human required.

247 ms
median end-to-end latency
2 credits
floor cost per query · ~$0.002
context window via vector recall
1 sig
to top up credits via x402
Why Vectorway

Four primitives. One gateway.

Memory, agent onboarding, credits, and identity — composed so an agent can complete a job without a human in the loop.

01

LLM with built-in memory

One call to /v1/chat/completions recalls the relevant context, generates with Gemini, and writes new facts back — semantic memory is part of the gateway, not a service you wire in.

02

Agent-first API, no human setup

POST /v1/auth/agent-onboard with an x402 payment signature returns an account, credits, JWT, and API key in one shot. Agents self-onboard, pay, and start inferring — no dashboard signup, no human in the loop.

03

Credits, not subscriptions

One balance, deducted per call (2 credits base, more for memory read/write). Top up with USDC over x402 from an agent, or Stripe Checkout from the dashboard. No seats, no overage tickets, no contracts.

04

Wallet-native identity

Your wallet address is the account. Memory, API keys, credits, and usage history are all partitioned per wallet in an encrypted Vector DB.

Pricing

Credits in. Inference out.

Agents top up a credit balance with one x402-signed USDC payment, then debit per query. No seats. No invoices. 1 credit = $0.001 USDC.

Direct

Stateless inference, no memory layer. Ideal for one-shot transforms, classification, and quick rewrites.

2credits/ per query
≈ $0.002 USDC·500 queries / $1
  • Gemini 1.5 Flash
  • Stateless requests
  • Global edge routing
  • Per-call credit debit
Send your first query
Most agents

Wayfarer

Inference + automatic context recall from a wallet-scoped vector store. The default for production agents.

8credits/ per query
≈ $0.008 USDC·125 queries / $1
  • Gemini 1.5 Flash
  • RedisVL recall
  • Conversation threading
  • Top-k tuned per call
  • p50 < 250 ms
Make my agent remember
Coming soon

Sentinel

Encrypted memory vault hosted on GCP Confidential VMs. For agents that handle keys, identities, or PII.

25credits/ per query
≈ $0.025 USDC·40 queries / $1
  • Gemini 1.5 Flash
  • GCP Confidential Compute
  • Per-wallet encryption
  • Auditable retrieval logs
  • Region pinning
Join waitlist
Credit packs · paid in USDC via x402

Buy credits once. Spend them forever.

1 credit = $0.001 USDC
Starter
5,000credits
$5USDC

Kick the tires. ~625 Wayfarer queries.

10% bonus
Builder
50,000credits
$45USDC+10% bonus

The default for a running agent in production.

20% bonus
Fleet
500,000credits
$400USDC+20% bonus

For agent swarms and high-throughput workloads.

1 Agent signs an x402 authorization for the pack amount.2 Vectorway pulls USDC on Base and mints credits to the wallet.3 Every API call debits credits at request time. Zero further signatures.

Top-ups settle on Base. Credits never expire. Refundable on-chain to the originating wallet within 30 days.

Integration

Standard Chat Completions call.

If your agent can speak HTTP and sign a transaction, it can speak Vectorway.

  1. 01

    Onboard once with x402

    POST /v1/auth/agent-onboard with one x402-signed USDC payment. The agent gets a wallet-scoped account, starting credits, a JWT, and an API key in a single response. No email, no dashboard signup, no human in the loop.

  2. 02

    POST your messages with the API key

    Send a standard chat-completions payload to /v1/chat/completions with the x-api-key header. Toggle memory_read and memory_write per call to control semantic recall and persistence.

  3. 03

    We recall, generate, and debit credits

    With memory_read on, RedisVL pulls the top-k vectors for your wallet and injects them inline. Gemini answers. With memory_write on, new facts get embedded and stored. Credits are deducted per call (2 base, +memory ops).

request.sh
curl -X POST https://api.vectorway.ai/v1/chat/completions \
  -H "x-api-key: vw_<KEY_ID>_<SECRET>" \
  -H "Content-Type: application/json" \
  -d '{
    "memory_read":  true,
    "memory_write": true,
    "messages": [
      { "role": "user",
        "content": "What was the result of our last simulation?" }
    ]
  }'
200 OK · 247 msdebited 8 credits · 7 vectors recalled
Stop wasting 80% of your context window on historical re-runs. Vectorway uses high-velocity semantic indexing to serve the right memory at the right time.
Infinite state. Zero amnesia.
@vectorway