GMO Pepabo began offering Lolipop! AI Gateway by GMO Pepabo, a managed AI gateway that lets developers call multiple AI models through one API key, on August 25, 2026. It provides unified API access to models from 14 providers including Anthropic, OpenAI and Google, and can automatically route each request to whichever of the user's selected models carries the lowest token price. All costs are shown in Japanese yen, and there is no setup fee or monthly base charge.

Swap the Endpoint and the Key, and It Runs

The API issued by the service is compatible with both the Anthropic and OpenAI specifications. Applications and tools already built against either of those APIs can start using the gateway simply by replacing the endpoint and the API key. No rewrite is required.

The supported providers are Anthropic, Google, OpenAI, Amazon, Meta, NVIDIA, Alibaba, DeepSeek, MiniMax, Mistral, Moonshot AI, Writer, xAI and Z.AI. The work of adapting to each provider's own API conventions, and of juggling separate contracts, is absorbed by the gateway.

Automatic Selection of the Cheapest Model

The headline feature is automatic selection of the lowest-cost model. Users nominate a set of candidate models in advance, and the gateway routes each request to the one with the lowest combined input and output token price. Preferred models, and fallbacks for when a chosen model is unavailable, can be configured per project.

Token prices differ by several times from one model to another even for comparable work. Deciding which model should handle which class of request has, until now, been a manual judgment on the development side. Turning that into a setting should cut down the operational fiddling.

Spending Visible in Yen

A dashboard shows token consumption, the model used and the cost for each request in real time, with every figure denominated in Japanese yen. AI gateways have been widespread overseas for some time, but GMO Pepabo cites the near-absence of domestic services offering yen-based cost management and Japanese-language support as its reason for entering the market.

Guardrails for team use are also in place. API keys, budgets and usage rules can be separated by project, and each API key can carry a monthly credit cap, a whitelist of permitted models and a per-minute request limit. That makes it possible to stop the familiar accident where one person keeps hammering an expensive model and the bill spikes at month end.

BYOK for Existing Contracts

Organizations that already hold direct contracts with AI providers can register their own API keys and route through the gateway under a BYOK (Bring Your Own Key) arrangement. In that case model usage is billed directly to the user's own account, and the gateway supplies only project management, cost visibility and guardrails. Registered API keys are stored encrypted.

On the enterprise side, the service also covers logging of requests and responses in addition to encrypted key management.

Prepaid Credits on a Usage Basis

Pricing has no setup fee and no monthly base charge. Users buy prepaid credits and draw them down against usage. Model charges follow each provider's own published pricing, and a platform fee of 5 percent applies when credits are purchased.

A semantic cache is planned before the end of 2026. When requests with the same intent arrive repeatedly, past responses are served from cache to reduce token consumption, with matching based on semantic similarity rather than exact wording. Further ahead, GMO Pepabo is considering additional models, a postpaid plan tied to usage volume, and an SLA-backed plan for enterprises.

A Fourth Piece After Run, Ship and Connect

GMO Pepabo has released a run of domestic infrastructure aimed at AI users and AI agents over the past several months. Lolipop! AI Agent Cloud by GMO Pepabo, for running AI agents, arrived in April 2026. Lolipop! Deploy Now by GMO Pepabo, for publishing web applications, followed in July, as did Lolipop! Zero Trust Link by GMO Pepabo for connecting devices and data.

The AI gateway adds a fourth piece, use, to that lineup. The company says it intends to integrate all four functions, run, ship, connect and use, into a single management layer so that domestic businesses and developers can work with AI while retaining data sovereignty.

Summary

Lolipop! AI Gateway by GMO Pepabo bundles 14 providers behind one API key and combines automatic cheapest-model routing, yen-denominated cost visibility and team-level budget control. With no setup fee, no monthly base charge and only a 5 percent fee on credit purchases, the barrier to trying it is low, particularly for teams already struggling to manage a mix of models and the spending that comes with them.