Infrastructure

API relay / token reselling

ActivePublished: 30 July 2026Updated: 30 July 2026Published

Key innovation

Reselling and redistribution of access to paid AI-model APIs through intermediary relay endpoints (中转), which decouple the end user from a direct account with the model provider.

How it works

1) The intermediary accumulates provider access: accounts, API keys or credits (own, bulk, or from pools/accounts). 2) It runs a gateway/relay (e.g. one-api, New API) exposing an OpenAI-compatible API and a billing, quota and key-management panel. 3) The client points base_url at the relay and uses the intermediary's key; requests are proxied to the chosen provider, with load balancing and key rotation. 4) The intermediary charges (often in tokens/credits) with a markup or discount versus original prices; payment via local methods.

Problem solved

Lowers the cost and access barriers to paid AI APIs (payments, regional availability, many accounts), offering one interface to many providers. The trade-off is legality, privacy and reliability.

Components

Relay / proxy endpointRequest forwarding

A forwarding endpoint that relays requests from the client's base_url to providers' original APIs.

Official

Key / account poolAccess source

A set of provider API keys or accounts used by the relay (with rotation).

Official

Billing and quota layerMonetization and control

A panel (e.g. one-api / New API) for managing users, tokens, quotas and load balancing.

Official

Payment integrationCharging

Local payment methods (Alipay, WeChat) and sale of credits/tokens.

Official

Implementation

Reference implementations

one-api (songquanpeng)

songquanpeng (JustSong)

OpenRouter (legitimate aggregation)

OpenRouter

Official

Implementation pitfalls

ToS violations and account bansHigh

Reselling usually breaks providers' terms; accounts/keys get banned, breaking the service.

Fix:Use legitimate aggregators (e.g. OpenRouter) and your own provider agreements.

Data leakage and prompt loggingHigh

The relay sees request and response content; the intermediary may log or exploit it.

Fix:Do not send sensitive data; prefer trusted, auditable providers.

Reliability and lack of SLAMedium

No direct provider accountability; quality and availability can be unstable.

Fix:Monitoring, failover, a fallback plan for direct access.

Fraud and model swappingHigh

Credits from abuse/fraud, and model swapping (bait-and-switch) to a cheaper model than advertised.

Fix:Model verification (fingerprinting), auditable billing, trusted intermediaries.

Evolution

2023

Rise of gateways and relays (one-api)

Inflection point

Open-source tools such as one-api make it easy to stand up relays / reselling stations for LLM access.

2024

Legitimate aggregation (OpenRouter)

OpenRouter and similar services normalize aggregating many models behind one API - the legitimate end of this market.

2026

Established gray market

Relay/proxy becomes a real infrastructure layer; in parallel, abuse grows (account pools, fraud, prompt logging).

Hyperparameters (configurable axes)

Markup / discountHigh

Price relative to original rates (markup or 30-90% discount).

Provider poolHigh

Which providers and models are aggregated behind one API.

Key rotation / poolingCritical

Strategy for rotating and failing over keys/accounts.

Computational complexity

Time complexity: O(1) narzut proxy na zadanie. Space complexity: O(uzytkownicy + klucze + logi).

Compute bottleneck

Provider rate limits and key-pool health

Relay throughput is limited by providers' aggregate rate limits and the number/validity of keys in the pool; account bans and credit exhaustion break the service.

Depends on

Kondycja puli kluczyLimity zapytan dostawcy