Engineering posts, research notes, and the Build your Agent with Abstraxn series featuring cloneable agents for research, trading, monitoring, and policy guardrails.
Latest
01AI agents can already run the supplier payment run. What finance teams lack is authority: what the agent may pay, who signs above the line, and proof an auditor accepts.
02The 12 questions enterprise buyers now ask in an AI agent security review, what each one is testing, and what a passing answer sounds like. Eight you can answer. Four you have to have.
03Know Your Intent (KYI) verifies what an AI agent is about to do, not just who it is. Why KYA repeats KYC, and why agents still have no equivalent of KYT.
04India already has a law that says an agent's purchase is the user's act. What it does not have is a way to prove what the user authorised. Every mandate scheme so far is operated by a party with an interest in the outcome. That is the gap.
05No API key, no monthly plan, just a wallet and a signature. Here's how we wired an agent that pays CoinMarketCap a few cents in USDC every time it asks a question, and what our first real test against an empty wallet actually looked like.
06An AI transaction simulation agent traces every transaction ID back to a real answer: confirmed or not, would it succeed or not. This agent previews outcomes before signing and decodes failures after, with Tenderly doing the simulation work underneath.
07A price check and a balance check are the easy part. Two guardrails run before Coinbase ever sees an order, and the agent never touches your real key. Here's how we wired it, bugs and all.
08Most research agents cite URLs they never opened. Here's how we built one that scrapes first, summarizes second, and ships from a repo you can clone in minutes.
09Before you ship an AI research agent or a trading bot, prove the wallet works. A five-minute smoke test for Agent Kit, MCP, and your LLM loop.
Builder reflections, market theses, and foundational thinking on the machine economy.
01Agent reputation registries are live and the data inside them is unusable. Portable reputation without portable enforcement is a category error.
02Agent payment rails inherited finality without reversal. Policy answers whether an action is permitted. Nothing answers whether it can be undone.
03On-chain agent policy failed on one property: readability. Once enforcement and legibility separate, the rule stays private and only the proof settles on-chain.
04On-chain, wallet, and application policy all fail autonomous agents. The enforcement position that matters sits above custody and below the application.
05Static credentials cannot govern autonomous agents. AI agent identity is shifting from a one-time signature to continuous, real-time validation of behavior.
06Mastercard, OpenAI, Visa, and MetaMask just validated the machine economy thesis in a single quarter. A builder's reflection on what they confirmed, what they missed, and why the interoperable layer is the gap that decides the next year of agentic infrastructure.
Know Your Intent: the practice of verifying what an AI agent is permitted to do before it acts, and producing a receipt anyone can check.
01AI agents can already run the supplier payment run. What finance teams lack is authority: what the agent may pay, who signs above the line, and proof an auditor accepts.
02The 12 questions enterprise buyers now ask in an AI agent security review, what each one is testing, and what a passing answer sounds like. Eight you can answer. Four you have to have.
Hands-on guides and open-source examples to help you clone, customize, and ship agents with Abstraxn Agent Kit.
01No API key, no monthly plan, just a wallet and a signature. Here's how we wired an agent that pays CoinMarketCap a few cents in USDC every time it asks a question, and what our first real test against an empty wallet actually looked like.
02An AI transaction simulation agent traces every transaction ID back to a real answer: confirmed or not, would it succeed or not. This agent previews outcomes before signing and decodes failures after, with Tenderly doing the simulation work underneath.
03A price check and a balance check are the easy part. Two guardrails run before Coinbase ever sees an order, and the agent never touches your real key. Here's how we wired it, bugs and all.
04Most research agents cite URLs they never opened. Here's how we built one that scrapes first, summarizes second, and ships from a repo you can clone in minutes.
Identity, wallets, payments, and policy primitives for autonomous agent systems.
01India already has a law that says an agent's purchase is the user's act. What it does not have is a way to prove what the user authorised. Every mandate scheme so far is operated by a party with an interest in the outcome. That is the gap.
02Cloudflare made any resource chargeable over x402. Here's the buyer stack an AI agent needs (identity, wallet, payments, and policy) to clear that paywall safely.
03How AI agents prove identity in agentic payments. Visa, Mastercard, Nuvei, AP2 and ERC-8004 compared, and why a portable identity layer beats siloed registries.
04Agentic commerce went live across Visa, Adyen, Stripe and Microsoft in a single week. The models were never the hard part. The agentic commerce infrastructure underneath is. Here is what the rails actually need.
05ERC-8004 is the on-chain identity standard for AI agents. Learn what agentic identity is, why an EVM address isn't enough, and how to register with the @abstraxn/agent-kit SDK.
06A spending limit in your agent's prompt is a suggestion. Here's how to build an AI agent spending mandate enforced at the wallet and x402 rail, so it can't be overspent.
07Everyone is racing to let AI agents check out. The winners will be whoever can prove what an agent bought, why, and on whose authority. A practitioner's guide to agentic commerce compliance, why it is the growth layer and not the friction, and the controls every autonomous transaction needs.
08Most agents silently fail when they hit a 402. Here's how to build one that detects the error, pays the API autonomously with USDC, and keeps going using @abstraxn/agent-kit.
09HTTP 402 Payment Required means your request was valid but payment is needed. Learn what causes it, how to fix it on npm, Stripe, WordPress, nginx, Azure, and why AI agents are finally making it relevant.
10How to give an AI agent a crypto wallet it can't drain: an on-chain identity, gasless x402 payments, and a hard spending cap, all via account abstraction.
11Two of the four powers are live. Here is how to give an AI agent an ERC-8004 on-chain identity and x402 payments with the Abstraxn Agent Kit, step by step.
12MetaMask just launched an Agent Wallet for AI agents. Here is what it signals for the agent economy, and how to build your own agentic apps with Abstraxn's ERC-8004 identity, x402 payments, and account abstraction.
13A bot follows rules, an assistant recommends, an agent acts. And on-chain, an agent looks exactly like a wallet. Here is what actually makes something an agent.
14Agent fraud is the easy problem. The hard one is the authorized, in-mandate transaction a user regrets. Why agent payment liability is a design choice.
15AI agent infrastructure needs four primitives: identity, wallet, payments, policy. Why ERC-4337 and ERC-8004 are the foundation.
Embedded wallets, recovery, and onboarding without seed phrases.
01Seed phrases give you ownership, not access continuity. Recovery has to be enforced by infrastructure, not memorized by users.
02Standalone bundlers process UserOperations. They do not solve execution. The unified-stack argument matters more than the bundler itself.
