EmbedPay vs OpenAI Embeddings (direct)
EmbedPay is an x402 wrapper around the same OpenAI text-embedding-3-small
model. Direct OpenAI is cheaper per token but requires an account, API key, and rate-limit
management. EmbedPay's premium buys signup-free access with composable x402 billing.
At a glance
| EmbedPay | OpenAI Embeddings (direct) | |
|---|---|---|
| Underlying model | text-embedding-3-small (1536d) | Same — text-embedding-3-small (1536d) |
| Signup required | No | Yes — OpenAI account + billing card + API key |
| Per 1k tokens (standard) | $0.00005 USDC | $0.00002 USD |
| Per 1k tokens (batch ≥100 inputs) | $0.00003 USDC | n/a — batch API has separate quota |
| Premium vs direct | ~2.5× direct rate. The premium covers x402 infrastructure + no signup convenience. | |
| Token counting transparency | cl100k_base counted server-side, returned in input_tokens field — you can verify locally | Same tokenizer, count returned in usage block |
| Rate limits | 600 req/min per IP · 10M tokens/IP/hour | Account-tier based (3k–10k+ RPM) |
| Model version pinning | Returned in response: model_version: "text-embedding-3-small (2024-01-25)" | Implicit — OpenAI rev-locks behind the scenes |
| MCP integration | Native via @melis-ai/x402-tools-mcp | Via OpenAI SDK wrappers |
| Multi-tenancy | Per-agent pay-per-call — no shared account | One account = one billing relationship |
Cost math at three volumes
| Volume | EmbedPay | OpenAI direct | Premium |
|---|---|---|---|
| 1M tokens | $0.05 | $0.02 | +$0.03 |
| 100M tokens | $5.00 | $2.00 | +$3.00 |
| 1B tokens | $50.00 | $20.00 | +$30.00 |
| 1B tokens, batch tier | $30.00 | $20.00 | +$10.00 |
At ~1B tokens you're $30 better off direct — and at that scale you probably want the dashboard, rate-limit insights, and support contract that come with an OpenAI org. EmbedPay's sweet spot is below ~100M tokens/month where the operational overhead of provisioning + securing an OpenAI key outweighs the price delta.
Choose EmbedPay when
- You're building an agent that needs embeddings but you don't want to manage an OpenAI key.
- Multi-tenant agent runtime where customers shouldn't share an OpenAI billing relationship.
- You're already in the x402 ecosystem — same wallet pays for ScrapePay → MarkdownOpt → EmbedPay → MemoryServe.
- Total volume under ~100M tokens/month and the wrapper premium is < engineering time saved.
- Need honest token-counting in the response for billing reconciliation across agents.
Choose OpenAI direct when
- High and stable volume — 1B+ tokens/month makes the per-token delta material.
- You already have an OpenAI account with negotiated enterprise rates.
- You need OpenAI's other embedding models (text-embedding-3-large at 3072d, ada-002) and want a single provider for all of them.
- You want dashboard analytics, fine-grained rate-limit visibility, and a support contract.
Try EmbedPay
Install once: npx @melis-ai/x402-tools-mcp. Then call the embedpay
tool from your agent. See the EmbedPay service page
for the full schema and the canonical RAG pipeline for
a worked composition example.