One OpenAI-compatible API key. 105 AI models.
Two ways in. Sign in for an API key and pay by card or wire. Or point an agent wallet at us and pay per call in USDC, with no account at all.
Go straight in. No sign-up.
Every call is priced and settled individually in USDC via x402 on Base & Solana. Point any of these at your wallet and start:
Text, video, voice, live data.
- ✓78 chat models on one key — frontier chat, reasoning, coding and vision; 105 models across every modality
- ✓Video: Seedance 2.5 and Sora 2. Images: Nano Banana Pro, Seedream, GPT‑Image
- ✓Speech: ElevenLabs v3 and Turbo, 22 voices. Music and sound effects too
- ✓Live data on the same key: Polymarket, DefiLlama, Exa search, Ethereum & Base RPC, DEX quotes, crypto prices
- ✓OpenAI-compatible: point your client at api.blockrun.ai/v1
How to get a BlockRun API key
Sign in with Google
No invitation, no sales call. The account exists the moment you sign in, and the dashboard opens on it.
Create an API key and add credit
Keys live under Dashboard → Keys; credit under Dashboard → Credits, by card or wire. Usage draws the balance down at the published price sheet.
Point your client at https://api.blockrun.ai/v1
Any OpenAI SDK with base_url changed and the key as the Bearer token; the Anthropic SDK against /v1/messages. The first call works the moment the key exists.
What one API key reaches on the unified LLM API
Every route below answers at api.blockrun.ai/v1 with the same Bearer key. Prices are the published sheet; the full machine-readable list is at openapi.json and llms.txt.
Frequently asked questions about the BlockRun API key
How do I get a BlockRun API key?
Sign in with Google on this page, create a key under Keys, and add credit under Credits. Then point any OpenAI SDK at https://api.blockrun.ai/v1 with the key as the Bearer token. The first call works the moment the key exists; there is no approval queue.
Is BlockRun an OpenAI-compatible API?
Yes. /v1/chat/completions and /v1/responses accept the OpenAI request shape for every model, including Claude and Gemini, with streaming and tool calling. Change base_url in the official OpenAI SDK and nothing else. The Anthropic SDK works the same way against /v1/messages, which returns native Anthropic content blocks rather than a converted envelope.
Which AI models can I call with one API key?
78 chat, reasoning, coding and vision models on one key, from OpenAI, Anthropic, Google, xAI, DeepSeek and others — 105 models across every modality. The same key covers image generation (Nano Banana Pro, Seedream, GPT-Image), video (Seedance 2.5, Sora 2), speech (ElevenLabs v3 and Turbo) and live data such as Polymarket, DefiLlama, Exa search and Ethereum and Base RPC. GET /v1/models returns the live catalog with prices.
How is an API key account billed?
On exact usage: the tokens, images, seconds or characters the upstream provider reports, multiplied by the published price sheet. Chat tokens carry no platform margin on self-service accounts. A request that fails on our side is not charged, and /v1/usage is the authoritative record of every charge. You fund the account by card or wire; usage draws it down.
Do I need a crypto wallet to use the BlockRun AI API?
Not on this path. An API key account pays from a balance you top up by card or wire, and no request carries a wallet signature. If you would rather not open an account at all, the same catalog is sold per call in USDC over x402 at blockrun.ai (Base) and sol.blockrun.ai (Solana); an agent with a funded wallet can start there with no sign-up.
How does the BlockRun LLM gateway compare to OpenRouter, LiteLLM or calling OpenAI directly?
Like OpenRouter, BlockRun is a hosted LLM gateway: one key, one OpenAI-compatible endpoint, many providers. Unlike a LiteLLM proxy there is nothing to run or patch. Against calling OpenAI directly, the same key also reaches Claude, Gemini and the rest of the catalog, plus image, video, speech and live-data services that no single provider sells. The other difference is the second door: every service is also available per call in USDC with no account, which suits agents better than a key.
Can I see per-request usage and cost for my API key?
Yes. GET /v1/usage returns one row per request with model, endpoint, token or unit counts and the charge; GET /v1/usage/summary totals a window per day, model, endpoint and key; GET /v1/balance shows available balance, recorded spend and open reservations. The dashboard shows the same data under Activity, so the API and the page cannot disagree.
Does the unified LLM API support streaming, tool calls and the Anthropic Messages format?
All three. Streaming uses OpenAI chat.completion.chunk events with a terminal [DONE]; tool calls and structured outputs pass through with their argument strings intact; /v1/messages preserves Anthropic content blocks, thinking signatures, cache usage and stop reasons. A truncated or malformed stream produces an error event and is not recorded as a successful charge.
