Synthorai
Sign up free, 10 calls on us. Up to $1, no card needed. DeepSeek V4.1 Flash - 65% OFF Input, output and cached input. Eligibility applies.

API Reference

Synthorai exposes a multi-protocol REST API over HTTPS. The same base URL, key, and billing serve three request shapes - OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages - plus auxiliary endpoints for embeddings, images, audio, and reranking.

Base URL:  https://synthorai.io/v1

All endpoints require:
  Content-Type:  application/json
  Authorization: Bearer <your-api-key>
  (or x-api-key: <your-api-key> for /v1/messages)

Endpoints

EndpointTypeDescription
POST /v1/chat/completions chat OpenAI Chat Completions - universal entry point, routes to any provider.
POST /v1/responses chat OpenAI Responses - tool use, reasoning, structured output.
POST /v1/messages chat Anthropic Messages - Claude / Bedrock-Anthropic native.
POST /v1/embeddings embedding Generate vector embeddings.
POST /v1/images/generations image Generate images from text prompts.
POST /v1/videos video Create and poll asynchronous video generation jobs.
POST /v1/audio/transcriptions audio Transcribe audio to text.
POST /v1/audio/speech audio Convert text to speech.
GET /v1/models read List all models available to the authenticated key.
/api/v1/keys keys Create, list, change, rotate and delete inference keys from your own code (requires an admin key with keys.read or keys.write).
GET /api/v1/requests logs List and read the workspace’s requests: model, key, tokens, cost, timing and labels, never the prompt or the answer (requires an admin key with logs.read).
/api/v1/prompts prompts Read and change the prompt library from your own code: prompts, versions, labels and rollouts, plus test sets and eval runs (requires an admin key with prompts.read or prompts.write; evals.read or evals.write for the tests).
GET /api/v1/audit-logs workspace Read how the workspace is set up and what was changed in it: the audit log, members and invitations, BYOK settings and export destinations, each with its own read scope. Credentials are never returned.
GET /api/v1/billing/* billing Read-only usage, cost, and freshness data for your own workspace (requires a billing key, not a regular API key).
POST /v1/mcp mcp MCP server: every model, API key management and billing data as tools for AI agents (Claude Code, Cursor, Codex …), with your API key.

Every endpoint bills against the same per-token catalog - see the model price comparison for what each routable model costs.