Skip to content
saygm — Subnets — AI — News

SayGM Adds OAuth Login for AI Apps With User-Funded Inference

The Bittensor gateway lets users authorize model access with separate permissions and spending limits for each app.

Table of Contents

SayGM, the AI inference gateway on Bittensor Subnet 28, has launched Sign in with SayGM, letting developers connect their apps to a user's existing AI balance through an authorization button rather than a manually pasted API key.

The feature keeps model usage on the user's bill while giving each connected app its own access credentials. Users decide which permissions to grant, set a monthly spending limit, and can withdraw access from their SayGM account. Developers can offer model-powered features without advancing the money for those users' inference, the process of running a model to generate a response.

The launch also connects app onboarding to SayGM's affiliate program. Developers who introduce new paying customers through the button can receive rewards in SN28 alpha, the subnet's token, from a shared customer-acquisition pool.

Separate Permissions and Budgets for Each App

A SayGM balance covers Claude, GPT, Gemini, and a selection of open-weight models. Connecting an app authorizes it to use that account within the permissions the user approves, avoiding the need to create and transfer a key by hand.

Apps can request three permissions. Use models allows requests with a dedicated key assigned to that app rather than the user's main key. Read usage lets the app see how much it has spent on the user's account. Read profile gives access to basic account details.

When granting model access, the user chooses a monthly budget or allows unlimited spending. Each app has a separate limit, only the user can change it, and the limits reset at the beginning of each calendar month in UTC.

The controls address a practical problem with sharing credentials among AI tools, where spending and access become harder to manage when several applications use the same key. A separate credential and budget make it possible to manage one connection without replacing the key used elsewhere.

Under Account → Connected apps, users can review each application's permissions and spending for the month. Revoking a connection immediately disables that app's access and its key.

OAuth Authorization Without a Shared Client Secret

The integration uses OAuth 2.0 with PKCE, short for Proof Key for Code Exchange. The app redirects the user to SayGM for approval, then exchanges the returned authorization code for its dedicated key.

PKCE ties that exchange to a randomly generated verifier held by the app that started the flow. The app sends a derived challenge during authorization and supplies the verifier when redeeming the code. Someone who intercepts the code alone cannot redeem it without the matching verifier.

That mechanism suits single-page web applications and mobile apps, where a fixed client secret embedded in the software cannot reliably remain confidential. Users approve access through SayGM instead of copying credentials between dashboards and application settings.

Once connected, an app can use SayGM's native APIs for the supported model providers. Each user pays for requests from their own balance at SayGM's published rates.

The launch follows TokenGauge naming SayGM its 2026 Provider of the Year for LLM API pricing, as previously reported by Intelligence. SayGM also routes requests through a hardware-sealed gateway; proprietary-model requests still reach the external model providers.

Referral Rewards Follow Qualifying Top-Ups

Developers can add an affiliate referral code to their Sign in with SayGM link. New accounts created through that button are attributed to the developer, which ties the app's signup flow to the gateway's existing rewards program.

SayGM advertises rewards worth up to 100% of qualifying referred payments, paid in SN28 alpha. Under its affiliate reward rules, eligibility runs for 30 days from a referred customer's first successful top-up, with attribution fixed at signup. Rewards are calculated on qualifying top-ups, not on the amount of AI inference consumed.

The program distributes a shared pool funded by the subnet's remaining miner emissions after SayGM's miner payouts. Rounds last at least one day, and each affiliate's share depends on its referred revenue relative to total qualifying affiliate revenue, subject to the reward cap and available pool. Unallocated alpha carries into later rounds, so the maximum advertised rate is not a fixed commission.

The alpha quantity is fixed when a round closes. Rewards for crypto-funded top-ups have a seven-day hold, while card-funded top-ups have a 120-day hold, both measured from that close. Refunds, disputes, and eligibility changes can reduce unpaid allocations; the token's dollar value can also change before payment.

Developers can obtain a referral link through the affiliates page and register an app through SayGM's contact form, selecting Partnerships as the topic.

Comments

Latest