By PressRelease
6 min read
Wyoming, USA, November 11th, 2026,
Apps and AI agents can now call open models served by GPU operators on Lightchain, pay a set fee per answer from a wallet, and trace each response to a job recorded on the blockchain.
Lightchain AI has opened its Developer API on its main network, giving developers a way to use open AI models, whose weights are published so anyone can run them, served by a network of machines instead of a single provider. The API is compatible with OpenAI's chat API, so an existing app can switch by changing the base URL its requests go to, the API key it sends and the model name.
That compatibility is the main pitch. Teams keep their OpenAI SDK and their request format, and can point one feature at Lightchain, compare the results, and move the rest only if it works.
How the API Works and What Developers Need to Know
The API follows OpenAI's chat completions format and works with the official OpenAI SDKs. Each call runs as one job on the Lightchain network: a worker serves it, the result is recorded on the blockchain (on chain), and the response names the job. Model names follow Lightchain's own ids, such as gemma4:e2b, and the live list of models that workers are serving is available from the API itself.
Lightchain's documentation is direct about what is different. The API is text only, with no images, tools or function calling, so agents that rely on tools will need changes. Settings such as temperature are accepted but have no effect, because each worker runs its model with the settings its operator configured.
Developers who want to test before spending real funds can start on Lightchain's test network, where a faucet hands out free test LCAI. A browser playground, on the test network and on mainnet, runs one request without an API key, using a wallet and some LCAI, then shows the on-chain job that answered it along with the same request as curl and SDK code. Moving from the test network to mainnet means changing the base URL.
For TypeScript, Lightchain publishes the @lightchainai/sdk package, which sets the right address for each network and can top up the balance on its own.
A Wallet Instead of a Sign-up Form
There is no account form. A developer signs in with a crypto wallet to mint an API key, then funds a prepaid balance with native LCAI in one transaction. Each answer is paid from that balance. Until the balance is funded, or when it can no longer cover a job, the API replies with the exact transaction needed, so a program can handle the top-up without a person stepping in.
The balance stays under the developer's control. The API can spend only within an allowance the wallet grants, only on AI jobs, and it cannot withdraw or move the funds. Developers can cap what each key may spend, revoke a key the moment it leaks, and withdraw the balance at any time.
For agents, Lightchain is also testing an experimental x402-inspired pay-per-call option. A program that has deposited LCAI once can attach a signed payment to each request, with no API key and no account. The option uses Lightchain's own payment scheme, so it works through the Lightchain SDK rather than off-the-shelf payment clients.
One fee per answer, known in advance
Because the fee does not depend on length, spending is easy to forecast: a developer can work out the cost of a thousand requests before the first one runs. Each job carries up to about 124 KiB of conversation, and each model has a set output limit. If no worker answers a request, its fee is refunded.
A Receipt with Every Answer
The network checks the work. A checking service called the disputer re-runs a sample of jobs and compares the results, and answers that do not hold up are recorded against the worker, which stakes 50,000 LCAI to join. Workers are assigned by a weighted random draw on the blockchain itself rather than by a central dispatcher.
The documentation is also explicit about the limits of the receipt. It proves that an answer came from a real, paid job served by a named worker. Because the API decrypts answers on its servers to keep the OpenAI format, developers cannot check the returned text against the encrypted answer committed on chain themselves, and the API's operator can read prompts and answers. Teams that need to keep prompts from the API's operator are pointed to Lightchain Chat or to the protocol directly, where the user holds the session keys.
Key Features and Tools Available to Developers
How Model Configuration, Pricing and Governance Work
Models and prices are settings stored on the blockchain, so every change is public. A multisig wallet controls them, and three of its five signers must approve a change. Holders of native LCAI vote on how treasury funds are used. A vote runs for about five days, and a proposal that passes waits 48 hours before it can be executed.
Getting Started
The quickstart guide at docs.lightchain.ai covers minting a key, funding the balance and making a first call.
About Lightchain AI
Lightchain AI is a layer-1 blockchain, a base network of its own, built to run AI. Workers serve open models on their own hardware, each answer is recorded as an on-chain job, and native LCAI holders vote on proposals through on-chain governance. The LCAI DUNA, a Wyoming association, is its legal association. Learn more at lightchain.ai.
This article is for informational purposes only and is not financial advice.
Lightchain AI media
lightchain@lunarstrategy.com
Decrypt-a-cookie
This website or its third-party tools use cookies. Cookie policy By clicking the accept button, you agree to the use of cookies.