Skip to main content
ByteSpike is a multi-model gateway that speaks Anthropic Messages as its native protocol, with OpenAI Chat Completions, OpenAI Responses, and Gemini Native shimmed transparently. Behind one key you get text, image, and video endpoints — billed in transparent credits, not per-vendor SKUs.

Quickstart

Make your first request in under two minutes.

Authentication

How API keys, group bindings, and rate limits work.

API Reference

Text, image, and video endpoints — one base URL, one auth header.

Pricing

Per-token / per-call rates, no markup tiers.

​
Why ByteSpike

  • Anthropic-compatible by default — keep your tool_use, cache_control, and thinking blocks. Same SDK, same retry semantics, every model.
  • Multimodal under one key — text, image, video — no per-vendor billing surface to assemble.
  • Failures don’t bill — every non-2xx is free. Estimated credits ship in the response header so you can preview cost before user confirmation.
  • Per-key controls — every API key carries its own quota (USD), rate-limit buckets (5h / 1d / 7d), IP allowlist/denylist, and optional expiry. Org wallets roll up across keys.

​
What’s behind the gateway

Three protocol surfaces, the full multimodal catalog, and a handful of utility endpoints — all served from llm.bytespike.ai: The live model list and per-model pricing are published by the gateway at GET /api/v1/model-catalog and rendered at bytespike.ai/models and bytespike.ai/pricing — always current, no hand-maintained list. This documentation focuses on the request shape, response shape, and gotchas for each.

​
Base URL

Anthropic SDKs work out of the box by setting baseURL to the value above. OpenAI SDKs work the same way — see Authentication for the per-protocol header layout.