Docs

Docs on API access, keys, billing, usage and error codes.

• Getting started

  From sign-up to your first call: introduction, quick start, authentication, and how model IDs, provider identifiers and endpoint identifiers work.

• Introduction

  One key for models over OpenAI-, Anthropic- and Gemini-compatible protocols, with several providers per model and the cheapest eligible endpoint by default.

• Quick start

  Sign up, verify your email, create an API key and make your first call through the OpenAI, Anthropic or Gemini compatible API, with sample code.

• Authentication

  How to send your API key over the three compatible protocols: Bearer, x-api-key, x-goog-api-key and the key query parameter, plus authentication error codes.

• Models and identifiers

  Model IDs, aliases, provider identifiers and route_ endpoint identifiers, and querying models with /v1/models and /api/v1/models.

• API keys

  How to create, pause and revoke API keys, why the full key is shown only once, and how to limit each key by allowed models, daily and monthly spend, IP allowlist and expiry.

• Model capabilities

  Which endpoints serve text, vision, embedding, image, audio and video models, how model status affects calls, and the no-silent-replacement rule.

• Apps

  Connect OpenAI-compatible apps such as Cherry Studio, Cursor and Dify: the API host, key and model settings, and what to check when they cannot connect.

• Routing

  How the platform chooses among providers for one model: the provider parameter, fallbacks, default routing, data policy and budgets.

• Provider selection

  Every field of the provider routing parameter, the header form, locking to one endpoint, validation, and three differences from OpenRouter.

• Fallbacks

  allow_fallbacks defaults to false: a failed first endpoint returns the error, and switching stays within one model without loosening constraints.

• Default routing

  Default routing: comparison-price order, even splits within 1% tiers, 5-minute session affinity via X-ModelPortal-Session, and what sort changes.

• Data policy

  Data processing regions of the platform, content moderation and providers, endpoint data policies, and how data_collection and data_regions constrain only providers.

• Budgets and limits

  Balance, daily and monthly key spend limits, rate-limit tiers and provider.max_price caps, with the error codes for each.

• Features

  Streaming, tool calling, structured output, prompt caching, reasoning, the Anthropic and Gemini APIs, and parameter mapping.

• Streaming

  Streaming across the three protocols, usage with include_usage, first-token and idle timeouts, and billing for interrupted streams.

• Tool calling

  Tool calling in the OpenAI, Anthropic and Gemini formats, routing only to endpoints with tool support, and session binding for signed thinking blocks.

• Structured output

  JSON mode and JSON Schema structured output in OpenAI Chat, Responses and Gemini, and the capabilities endpoints need.

• Prompt caching

  Anthropic cache_control and automatic upstream caching, cache read and write pricing, and why endpoints without caching never fail a request.

• Reasoning

  The unified reasoning parameter versus reasoning_effort, thinking and thinkingConfig, level mapping, and billing of reasoning tokens.

• Anthropic Messages API

  Connecting Claude Code and the Anthropic SDKs: the site URL as base URL, required max_tokens, routing preferences in a header, and the error format.

• Gemini API

  Calling the Gemini-compatible API with the Google GenAI SDK or REST: three auth options, model IDs in the path, streaming and the error format.

• Parameter mapping

  How temperature, top_p, top_k, seed, n, logprobs, max_tokens, tools, structured output, reasoning, caching and multimodal input map across protocols.

• Billing

  Prepaid billing, attempts and settlement, usage and billing records, and refunds.

• Billing

  Prepaid billing rules: paid balance and gift credit order, per-token, per-image and per-second prices, price version locking, free failures, low balance and arrears.

• Attempts and settlement

  How attempts are billed: you pay only for the successful one, failed attempts are free and unpaid to providers, and how balance holds work.

• Usage

  View requests, failure rate, spend and savings in the console, split usage by model and key, query and export billing lines, plus timezone and privacy notes.

• Refunds

  Refunds return only unspent top-ups; gift credit is not refundable, and requests need your password and a finance review.

• Help

  FAQ, error codes, status page and privacy.

• FAQ

  Common questions on base URLs, empty model lists, charges for failures, 402 with balance left, choosing providers and keeping sessions.

• Common errors

  Error formats of the OpenAI, Anthropic and Gemini compatible APIs, and causes and fixes for common codes such as invalid_api_key, insufficient_balance, rate_limited and model_not_found.

• Status page

  Where the status page gets availability, time to first token and throughput, the 5-minute probes, and incident and maintenance notices.

• Privacy

  The platform keeps no request or response content except for risk review, strips user-linking fields before forwarding, and providers follow their declared policies.

• API reference

  Addresses, authentication, request fields, response examples and error codes for the OpenAI-, Anthropic- and Gemini-compatible APIs and the models API.

• Overview

  Base URLs and versions, authentication, the X-ModelPortal-Provider-Preferences and X-ModelPortal-Session headers, errors and rate limits.

• OpenAI-compatible API

  Request fields, response examples and error codes for /v1/chat/completions, /v1/responses, /v1/embeddings and /v1/models.

• Anthropic-compatible API

  Request fields, billed usage, error types and x-should-retry for /v1/messages and /v1/messages/count_tokens.

• Gemini-compatible API

  Request fields, billed usage and error statuses for generateContent, streamGenerateContent, countTokens and embedContent.

• Models API

  /v1/models and the platform format /api/v1/models: query parameters, endpoint and provider identifiers, declared configuration and data policy fields.