Sabitlenmiş Tweet
Been thinking a lot about what it actually takes to build production onchain AI agents.
Not demos. Not toys. Real ones.
Here’s the framework I’ve been using: SPIDER
S - Security. Secrets, access control, LLM threat surface. Everything else collapses without this.
P - Payments. How agents pay each other. x402, MPC, A2A payments. Agents need to transact, not just talk.
I - Identity. DIDs, ENS, and other identity protocols. An agent without identity is just a script.
D - Discovery. How agents find each other. Google A2A, ERC-8004, and the emerging agent registry layer.
E - Ethereum. Where agents write and manage their own state, own contracts, their own keys or smart accounts. Debatable this could be another blockchain but Ethereum makes the most sense.
R - Runtime. Where the agent actually lives. Local to cloud to bare metal, and how it carries memory across that lifetime.
Most people building agents are nailing 1 or 2 of these.
Production means you need all 6.
What do you think? Tag the tool for each pillar you use below 👇🏼
English
