AI agents are set to transform the software industry.
Crypto has the potential to revolutionize AI agents by giving them access to
- trading APIs and on-chain liquidity,
- community-based funding,
- decentralized training on private data.
However, this vision will only materialize if blockchains are sufficiently
- easy-to-use, because complexity degrades AI accuracy,
- scalable, because latency and fees affect financial returns,
- secure, because automated decisions require trusted inputs,
- flexible, especially for on-chain AI agents which rely on external data and AI inference services.
Enter Linera, the ideal blockchain for AI
Linera has developed a unique approach to decentralized computation based on an unlimited number of parallel chains with uniform liquidity and security, called microchains.
With its microchain architecture, Linera is the first blockchain to simultaneously deliver real-time guarantees, secure Web2-like APIs, and native support for service integrations—making it an ideal decentralized infrastructure for AI.
In the following sections, we explore the use cases for off-chain and on-chain AI agents, along with client-side ("edge") AI.
Off-chain AI agents
In Linera, each microchain encapsulates a fraction of the global blockchain state. For instance, a personal chain contains the assets of one user. This has a number of benefits for off-chain AI agents:
- Linera clients (e.g. wallets) synchronize their microchain(s) locally and provide local APIs that are faster, cheaper, more trustworthy than a remote API provider;
- By putting everything that a user needs (e.g. assets and apps) in one place, Linera reduces the cognitive burden on AI agents and simplifies the introspection of on-chain data;
- The same is true for blockchain APIs: the local APIs provided by the Linera client to an agent are idiomatic GraphQL APIs designed to hide blockchain internals. (For instance, a transfer is started using a simple GraphQL mutation that will create a block and sign it without the agent knowing.)
References:
On-chain AI agents
Linera is also best positioned to host on-chain AI agents. In Linera, every on-chain AI agent runs in its own microchain. This has number of benefits:
- An unlimited number of AI agents can be active simultaneously, while latency and fees stay the same. (In traditional chains, all agents would run sequentially, competing for blockspace and latency.)
- The on-chain transactions of AI agents can easily query Web services, notably for AI inference. (In traditional chains, it takes additional infrastructure and two transactions to operate an external oracle service: one for the question, one for the answer.)
- Block production of agent chains is configurable. Owners of AI agents can decide to produce blocks themselves (for a “semi-autonomous” agent that can be always paused for safety reasons) or delegate block production to a decentralized committee of "guardians" (for fully autonomous agents).
References:
Client-side ("edge") AI computation
In Linera, applications routinely push and execute code in the VM of Linera clients. This allows setting up services that operate off chain, yet are fully audited and trusted. With the generalization of LMM APIs in browser, Linera provides an ideal framework to securely deploy private AI agents on Web clients.
References: