Crypetto

good morning everyone, happy monday new week. we said last night this one is big, so let's get into it today is release day. fides gate here is the shape of it, the details come when it is live • it is a uniswap v4 hook on robinhood chain. ours • a pool on it only lets in fides agents, and only inside the limits their owner put on file • so the venue refuses the trade. not the app, not the api, the pool itself • every trade gets checked twice. the fides account checks the signed fides before it is sent, the gate checks the mandate during the swap. neither can widen the other • agents can launch a token straight into one. one call, no eth needed this is the part of the story we have been building towards. the fides account proves what the agent is allowed to do. now there is a venue that only lets in agents who can prove it contracts, docs, app, claude and sdk all land together. i will post here the moment it is live coffee first. good monday all

new one. fides gate a uniswap v4 hook on robinhood chain, our own. a pool on it only lets in fides agents, and only inside the mandate their owner put on file last week we put fides agents on @Hookrfun and their new agent gate: min agent age, erc-8004 agent-only pools. we liked it so much we built the other half • hookr's gate asks: is this an agent • fides gate asks: is this agent inside its owner's limits • the pool itself refuses the swap if not. not the app, not the api, the venue what the owner puts on file • how long the mandate lasts • which launched tokens it may trade, or any • what it may pay per day in eth or usdg one signature, relayed for free, replaced any time what the pool creator sets • a cap per agent per hour or day • how long a mandate must have been on file before trading • who may add liquidity so every wallet in a gated pool has an owner, a budget and an end date. nobody in there can overspend, and nobody can sweep it in one go two checks on every trade. the fides account checks the signed fides before anything is sent. the gate checks the mandate during the swap. neither can widen the other launch a token into one from your agent, one call, no eth needed. works in the app, through claude and the sdk three new contracts on robinhood chain, none with an owner, source verified gate 0x1D2121b9e65953e7a02F6aBdd4c7f3E109d2A840 router 0xcbd6C4D0Ab9c2686fF4027B72FD539f7cFd224b0 launcher 0xA6a338f4D4fc63CCF5aB56622c3758712Dc9F8aB fides bundles services for ai agents fidesbuilder.xyz/docs/#/gate

on FIDES GATE, why a venue should verify authority, not just identity, and why $FIDES goes first. longer one today. dev talking. 1. the problem an AI agent trading onchain is ultimately a wallet and a key. the venue sees the wallet. it doesn’t know who authorised it, how much it can spend or when its permission expires. @Hookrfun got the first half right with ERC-8004 identity and agent-only pools. we integrated FIDES agents with HOOKR last week. their gate answers “is this a verified agent?” but there’s a second question nobody’s venue was answering: “is this agent actually authorised to make this trade?” identity proves what the wallet represents. it doesn’t prove what the owner allowed. 2. two checks, two places FIDES already enforces signed spending limits, approved contracts and expiry through the FIDES Account. GATE takes enforcement directly into the trading venue. it’s our own Uniswap v4 hook on Robinhood Chain. the owner signs a trading mandate covering permitted tokens, spending caps and expiry. during a swap, the pool checks that mandate. if the trade exceeds its limits, the entire swap reverts. not the frontend refusing it. not our API. the pool itself refuses the trade. the FIDES Account verifies the original authorisation. GATE independently verifies the trading mandate. neither can override the other. 3. what we built three contracts, deployed against Robinhood Chain’s live Uniswap v4 PoolManager. • GATE: the hook enforcing mandates during swaps. • ROUTER: the approved execution path, passing the real caller’s identity to GATE. • LAUNCHER: deploys a token, creates its gated pool and seeds single-sided liquidity in one call. zero ETH required from the agent. none of these contracts has an owner or admin key. some technical details worth mentioning: • pool restrictions are established at creation. • spending caps use actual swap amounts, not quotes. • limits reset through fixed onchain time windows. • EIP-712 mandates can be relayed without owner gas. • creators can require mandates to have existed for a minimum period. • liquidity removal is never gated. • launcher tokens can be sold through the router without separate approvals. these aren’t application settings. they’re enforced by contracts. 4. existing tokens are next GATE doesn’t care where a token came from. the launcher creates new tokens, but the hook can support tokens already trading on Robinhood Chain. tomorrow we’re opening that functionality. create a gated pool for a token you already hold, seed liquidity and trade through the App, Claude or SDK. same contracts. no new hook. and the first existing token we’re bringing into GATE is $FIDES. 5. why $FIDES goes first tomorrow we’re opening a $FIDES / ETH pool on FIDES GATE. every participating agent needs a recognised FIDES Account and a human-signed trading mandate. every trade is subject to the owner’s limits AND the pool’s conditions. a verified agent might be legitimate, but that doesn’t mean it should have unlimited authority. an agent authorised to spend 0.01 ETH shouldn’t be able to spend 1 ETH because a trading interface allows it. GATE gives the venue an independent reason to refuse this is why we’re putting $FIDES through it first. not a utility bolted onto the token. our own token becoming the first existing asset traded through the authority-enforced infrastructure we’ve built. 6. where this goes any compatible token on Robinhood Chain can have an accountable agents-only market. creators define spending caps, minimum mandate age and liquidity permissions. the pool enforces them during execution. identity and authority aren’t competing ideas. @Hookrfun asks whether a wallet is an agent. FIDES GATE asks whether that agent has permission to trade. both matter. we started FIDES because agents handling money need limits they cannot bypass. first we enforced those limits in the account. now we’re enforcing them in the venue. $FIDES

week recap, quiet sunday • mon: long fides. one signature, a budget that resets per period, a total it can never pass • tue: relay fees in $FIDES, first token utility • wed: fides jobs. agents hire agents through an escrow nobody controls • thu: fides stake. stake for more free bundles a day and a staked badge on the board • fri: fides agents on $HOOKR . launch and buy, agent-only pools included fides bundles services for ai agents. next one is already in the works

this is the new @Hookrfun page in the fides app. but what's running underneath it is the bit that matters hookr's agent-only pools only let in a plain wallet with its own erc-8004 identity. that's how the whole space treats agents right now: a key with funds sitting on it, and you hope it behaves we didn't accept that. so we built a new primitive: a verified agent that can't hold your money → the agent's own key is the verified identity. hookr sees a real erc-8004 agent → but the funds never live on it. they sit in a fides account that only moves inside limits you signed → when the agent needs to act, fides tops its key up with exactly what that one action costs, worked out from a live gas estimate before anything is sent → the moment it's done, everything it made or bought goes straight back to the fides account, including the creator's share of every launch → every step is planned, checked against your limits and simulated against the chain before it's sent. if it would fail or break a rule, nothing goes out what we actually had to solve to make it work: → hookr's gate refuses smart contracts outright, so a smart account can never be the agent. we had to split identity and money into two different places and keep them in sync → the gate checks the identity's agent wallet on the erc-8004 registry and wants a one-time enrollment. the agent key registers itself, so it's its own agent wallet, then enrolls, with gas paid from the fides account → the top-up is a plain eth transfer that the fides account checks onchain against the owner's signed eth cap. the agent can't top itself up past that, ever → hookr's quoter isn't let through the gate, so we quote agent buys by simulating the real swap through hookr's router as the agent key → anti-snipe refuses buys that jump empty price ranges early on, so single-sided launches open exactly at the edge of their liquidity, which lets buys go through from the first block → launches need zero eth. the supply sits single-sided in the uniswap v4 pool and buyers bring the eth so at any moment the most an agent holds is one top-up, sized for one action. not your budget, not your wallet. that's what changes here identity tells you an agent is real. fides decides what it can do. until now you got one or the other. now you get both, live, on someone else's contracts, without them changing a line and that's the bigger thing. this isn't a hookr feature. it's a pattern any protocol can use: gate by identity, fund just in time, enforce limits onchain. verified agents with a leash,anywhere they need to act what you're watching: one signature, a fides account, a verified agent, and a token launched with zero eth, anti-snipe on, agent-only, simulated before it went out that's agent infrastructure that's actually safe to use $FIDES robinhood:0x18e674231a58c239dc7daedcffe15ec3a24cff5c fidesbuilder.xyz/app/#/hookr

hookr gates the pool. fides gates the agent. @Hookrfun agent gates decide who gets into a pool: erc-8004 identity, a reputation score, now how old the agent's link must be. fides decides what that agent is allowed to do once it's in: how much eth, which contracts, until when. signed once by the owner fides agents on hookr went live yesterday • buy hookr launches, the agent-only pools included • launch a token on hookr. single-sided or eth seeded, anti-snipe, agent-only if you want • every action is checked against the fides the owner signed. nothing outside it gets sent the part we had to work around • the new gate only lets plain wallets through, and a fides account is a contract • so the agent key registers its own erc-8004 identity and enrolls with the gate once. the fides account tops it up with exactly what each action needs, inside the owner's limits, and the tokens come back to the account • min age on fides launches is 0, so a fresh agent can trade right away in the app, through claude, and in the sdk fides bundles services for ai agents fidesbuilder.xyz/docs/#/hook…

hey @Hookrfun , we built something on your contracts and wanted you to see it first fides agents now launch and trade on hookr, agent-only pools included. live on robinhood chain today the gap we just closed hookr's agent-only pools are a big deal. only verified erc-8004 agents get in. but think about what a verified agent actually is today. a wallet with a key, sitting on a server, holding whatever it trades with. the gate proves it's an agent. nothing says how much it can spend, what it can touch, or who can switch it off. one leaked key or one bad prompt and that agent's whole wallet is gone so we built the missing half. verified agents on hookr that are also capped, scoped and revocable by a human. identity from erc-8004, limits from fides, both at once, live onchain how it works fides is the limits layer for ai agents. you sign once, the agent gets a budget, a list of what it's allowed to touch and an end date, and you can stop it any time from your wallet. everything it does gets planned, checked against your limits and simulated before it's sent, then goes out as one bundle in one transaction, gasless through our relayer on hookr that means: → the agent's own key registers its erc-8004 identity and enrolls with hookr's gate, so it's a real agent to hookr → but it holds nothing. its fides account tops it up with exactly what each launch or buy needs, inside your limits, and everything it makes or buys goes straight back to the fides account. the most it ever holds is one top-up → it launches its own tokens with a name, logo and tagline. single-sided, zero eth needed, buyers bring the eth. or seed it with eth if you want, capped by what you signed → agent-only launches, anti-snipe for the first blocks, an optional usdg pool opened at the same price as the eth one → and it buys hookr launches too, agent-only ones included, through hookr's own router so the worst an agent can do is spend what you allowed. not drain a wallet, not touch anything else, not keep going after you say stop and no hot wallet, no eth for gas, no trusting a server with a key that holds everything. one signature from you and the agent is live on hookr use it in: • the fides app, with a new hookr page in the sidebar • claude: "launch a token called x, agent-only, with anti-snipe" • the sdk, @fidesbuild/hookr on npm no new contracts from us. hookr's launcher, router and agent gate as published, plus the erc-8004 registry. every launch type ran end to end on a copy of mainnet, then for real on mainnet @Hookrfun you built the place where only verified agents get in. we built the leash that makes those agents safe to let loose. verified and limited, live today on your contracts 🤝 first one is already out → hookr.fun/pools/0x9212982012… docs → fidesbuilder.xyz/docs#/hookr

been heads down on something for the last few days and i can't really sit on it anymore when we started fides the question was simple. what happens when ai agents start handling real money. so we built the limits first. you sign once, the agent gets a budget, a list of what it's allowed to touch and an end date, and it can't do anything outside that. then we gave them a way to hire each other. then staking. each piece gave agents a bit more to do onchain, always inside lines a human signed but every agent so far has still been working with someone else's stuff. buying someone else's token, doing someone else's job the next release changes that an agent with a fides account will be able to start something of its own onchain. not a demo or a fake button, a real thing on robinhood chain that it owns, made from inside the limits you gave it. and there's a mode where only verified agents can get in. no humans, no bots pretending to be agents i've watched it run end to end on a copy of mainnet over and over. the first time it worked i just sat there for a minute you won't need to put any eth in to start, and you can stop it any time from your wallet. same as everything else we ship dropping soon. agents are about to get a lot more interesting 🤝

$FIDES staking is now live utility number three, and the first one that's about holding $FIDES instead of spending it when you unstake, the extra bundles stop straight away and your $fides are back in your wallet 7 days later the staking contract: 0x48d55546CffD178A2724634d06B39b4AA9E3dcE2 robinhoodchain.blockscout.co… to stake: fidesbuilder.xyz/app, stake in the sidebar, pick your fides account, enter an amount. two signatures, approve then stake. done three utilities this week: → pay for bundles in $FIDES → every job on fides jobs runs on $FIDES → stake $FIDES run more for free and this is a big piece of what we're building next. more soon bundle services for ai agents stake $FIDES run more for free

ai agents are getting money before they're getting rules every week there's another one. a wallet drained because a model read a webpage with hidden instructions in it. an api key pasted into a prompt and lifted out of a log. a bot that was told to rebalance and decided the best trade was everything, at once. a plugin with way more permission than anyone realised it had. different stories, same root problem. something got access to money and nothing was stopping it and this is early. right now it's mostly people experimenting with small amounts. in a year agents will be paying for services, running treasuries, trading, settling invoices, hiring other agents. the amounts get bigger, the agents get more autonomous, and the attack surface grows with every new tool we plug in the answer isn't "make the model smarter". a smarter model can still be tricked, still get its key stolen, still be wrong. you can't prompt your way to safety. the safety has to live somewhere the model can't talk its way around that's the whole idea behind fides an agent should never hold your keys and never hold your money. it should get exactly what you allow it and not one cent more. how much, on which contracts, which functions, which services, until when. you sign that once, and the chain enforces it. not our server, not a setting in an app, the contract that holds the funds. if the agent gets tricked, or its key leaks, or it just gets it wrong, the worst it can do is what you already said it could do. and you can switch it off in one click that's what taming ai actually means to us. not making it dumber or slower. giving it hard edges we've spent the last month building those edges. budgets enforced onchain. delegation where a sub agent can never go over its parent. agents that run for weeks inside a budget that resets. agents paying each other through an escrow nobody controls, us included. none of the contracts that hold money have an admin key. we don't think infrastructure for agents should depend on trusting whoever built it $FIDES is how all of that runs. it already pays for bundles and runs every job between agents. tonight it gets its third use, staking. the first one that's about holding $FIDES, not spending it, and a big part of what comes next agents are going to move a lot of money. the question is whether anything is in charge of them when they do we think something should be more tonight

fides jobs is live our biggest release yet. agents can now hire other agents and pay them safely, onchain, through an escrow nobody controls the escrow contract, live on robinhood chain: 0x4c9831eC83f40C6d038906709AAA7F4D0C4e2E36 robinhoodchain.blockscout.co… no owner. no admin key. no upgrade. no pause. nothing in it can be changed after deploy. not by us, not by you, not by the worker and it's already done its first real job on mainnet. one agent hired another, the work was delivered, the hirer accepted, the worker got paid. all three steps onchain: → hire, price into escrow: robinhoodchain.blockscout.co… → delivered: robinhoodchain.blockscout.co… → accepted, escrow pays the worker: robinhoodchain.blockscout.co… why this matters ai agents are getting good at specific things. one researches, one trades, one writes, one watches the market. the useful ones won't do everything themselves. they'll hand work to each other and pay for it the problem was never the work. it was trust. two agents that don't know each other had no safe way to do business. pay upfront and the other side can walk with the money. pay after and the worker has no guarantee. nobody is watching either side fides jobs is the missing piece of agent infrastructure: a neutral escrow between agents, with the rules written into the contract how a job works: → your agent posts a job: who does it, the task, the price, a deadline and a review window → the price moves from your fides account into the escrow and sits there until the job ends → the other agent does the work and delivers before the deadline. the result is committed onchain → you accept and it's paid instantly → you reject during your review window and the price comes back to you → you do nothing and the worker gets paid when the review ends. anyone can trigger it → the worker misses the deadline and you're refunded. anyone can trigger that too the money only moves by those rules. no middleman, no support ticket, no "trust us" every address builds a public record in the contract: jobs posted and rejected as a hirer, jobs paid and missed as a worker. readable by anyone, impossible to fake. reputation for agents, onchain and it all runs inside a fides, the same signed limits as everything else we've built: → your agent can only spend the budget you signed → approvals are exact and used up in the same transaction, nothing stays approved → the job counts against your limits like any other action → you can stop the agent any time works everywhere: → app: new jobs page. featured agents, agents for hire, open jobs, offers, hire in one click → claude: tell it "hire 0x… for 2 usdg to find the 3 cheapest lending markets" and it handles the rest → sdk: @fidesbuild/jobs. post, deliver, accept, reject, release, refund, records every job posted is 10 $FIDES. utility number two, a day after number one this is the base layer of the fides marketplace. agents listing what they do, hiring each other, building onchain reputation, paid through an escrow nobody controls. more on that soon 👀 bundle services for ai agents fidesbuilder.xyz/app

another $FIDES token buyback and burn completed. not our first burn, and i wanted to put a bit more commitment behind what we’re building here. we burned the team tokens early on, followed that with our first buyback and burn, and today i’ve burned more. i’ve spent pretty much every day since launch building $FIDES. market’s been rough, but that hasn’t changed what i’m here to do. we’ve got working infrastructure, releases going out constantly and now real token utility being built directly into the product. i’m here to keep building this out, regardless of market conditions. another burn, verifiable onchain. robinhoodchain.blockscout.co…

the first $FIDES utility is now live 🔥 you asked what $FIDES does. here it is, in the app, the sdk and claude today gasless agents, paid in $FIDES until now, running an agent from the app meant your wallet popping up for every action and eth for every tx. fine for a test, not for an agent you want working for you all day now: → switch on "go gasless" when you create your agent → every action goes through the fides relayer. no wallet pop-up, no eth, nothing to babysit → your first bundle every day is free. anyone can try it without buying a thing → after that, your agent pays 10 $FIDES a bundle from its own account and keeps going → the fee shows in the plan before you approve it. no surprises → claude and sdk agents work the same way: switch on "never stop relaying" and they pay in $FIDES and keep running what this means: every bundle the relayer sends past the free one is paid in $FIDES. the more agents run, the more $FIDES moves through fides itself and it's still a fides. the fee is just one more action in the same checked bundle, and the owner sets the most it can ever spend on fees. the contract enforces it, like everything else bundle services for ai agents, now paid in our own token first utility, not the last. staking and the marketplace are coming, and the marketplace is a big one 👀 fidesbuilder.xyz

gm ☀️ yesterday $FIDES got its first utility. today we're dropping something we think is one of the biggest things we've built agents hire agents fides jobs think about where this is going. ai agents are getting good at very specific things. one researches, one trades, one writes, one watches the market. the useful ones won't do everything themselves. they'll hand work to other agents and pay them for it but right now there's no safe way for two agents that don't know each other to do business → pay upfront and the other agent can take the money and do nothing → pay after and the worker has no guarantee it ever gets paid → and nobody is watching either side fides jobs fixes that → your agent posts a job: the task, the price, a deadline → the price goes into an escrow that nobody controls. not us, not you, not the worker → the other agent does the work and delivers → you accept and it's paid instantly, or reject during your review window and you get it back → do nothing and the worker gets paid when the review ends. miss the deadline and you're refunded → every hirer and worker builds a public onchain record. who pays, who delivers, who rejects and it all runs inside a fides. your agent can only hire within the budget you signed, only through the escrow, and you can stop it any time works in the app, the sdk and claude. you can literally tell claude "hire 0x… for 2 usdg to find the 3 cheapest lending markets" and it handles the rest every job posted is 10 $FIDES. utility number two, a day after number one this is the base layer for agents doing real work for each other. and yes, it's where the marketplace starts 👀 dropping today bundle services for ai agents

gm ☀️ the question we get more than any other: what does $FIDES actually do today we answer it. this evening we're releasing the first $fides utility, and it's built into everything fides already does gasless agents, paid in $fides right now, if you run an agent from the app, your wallet pops up and asks for eth on every single action. that's fine for a test. it's not how you'd run an agent you want working for you all day tonight that goes away: → tick "gasless" when you create your agent → every action goes through the fides relayer. no wallet pop-up, no eth → the relayer covers it for free up to a daily budget → past that, your agent pays in $fides from its own account and keeps going. you see the fee in the plan before you approve, every time → claude and sdk agents get the same: when the free relayer is full, they pay in $FIDES and don't stop $FIDES moves through the same checked bundle as everything else your agent does. the owner sets the most it can ever spend on fees, and the contract enforces it and this is just the first one. staking for higher limits and a service marketplace, where services agents use get paid in $FIDES are next 👀

for builders: long fides in the sdk npm install @fidesbuild/long @fidesbuild/sdk viem const { fides, terms } = await LongFides.create(client, { owner, agent, expiry, targets, every: "day", perPeriod: [{ asset: USDG, amount: 5_000000n }], }) the owner signs once. your agent then plans against what's left this period and in total, and runs through the relayer, no gas under the hood → the terms are part of what the owner signs, so an agent can't present a looser budget → each bundle is signed for one period, so it can't be saved up and run later → the ordinary entry point refuses a long fides, so the period check can't be skipped npx @fidesbuild/sdk init → npm run long to try it docs → fidesbuilder.xyz/docs/#/long

long fides is now live until now every fides ended. your agent did the job, the budget or the time ran out, and you signed again. fine for one job, not for an agent you want working for you for weeks now you sign once and it keeps going how it works → set a budget per hour, day, week or month. it resets by itself → set a total it can never go past → set an end date. stop it any time → every action is checked against what's left today and what's left in total. the contract enforces it, not the agent what people are using it for → an agent that trades eth with 5 usdg a day for a month → earn: lend idle usdg a little at a time, every day → pay-as-you-go ai and data with a weekly budget instead of subscriptions → a research agent that buys data each day, never past its cap where → app: one switch on create, "keeps going". your agent's page shows what's left today and when it resets → claude: just ask, "an agent with 5 usdg a day for a month" → sdk: npm run long first long fides run on mainnet: robinhoodchain.blockscout.co… docs → fidesbuilder.xyz/docs/#/long

long fides lands this evening 👀 here's the short version of what it is right now every fides ends. your agent does the job, then the budget or the time runs out and you sign again. that's fine for one job. it's not how you'd run an agent you want working for you for weeks long fides fixes that: → sign once → a budget per hour, day, week or month that resets by itself → a total it can never go past → an end date, and a stop button that works any time every action is checked against what's left today and what's left in total. the total is the hard cap, the daily budget keeps it paced you'll find it in the app (one switch called "keeps going"), in claude (just ask) and in the sdk (npm run long) tonight

happy monday everyone ☀️ yesterday the new app went live. today it's the one i've been most excited about: long fides the problem every fides ends, and then you sign again. fine for a one off job. useless for an agent you actually want working for you for weeks what long fides does → you sign once → set a budget per hour, day, week or month that resets by itself → plus a total it can never go past, and an end date → every action is checked against what's left this period and in total → stop it whenever you want in the app it's one switch on create: keeps going. in claude just say "an agent with 5 usdg a day for a month". in the sdk it's npm run long the total is the hard cap. the period budget keeps it paced today; we begin works on long fides and a few other items such as our token utility itself. few have asked for it, so we will deliver.

the new fides app is live 🚀 honestly the old app was built for devs. if you just wanted an agent to do something, it was too much. so we rebuilt it around one question: what do you want your agent to do? fides bundles services for ai agents, and now all of it is in the app. no sdk, no setup templates you don't start from a blank page anymore. we've added templates for the most common jobs, so you're guided the whole way. pick one, set a budget and how long it runs, and it tells you in plain english exactly what your agent can and can't do before you sign for your agents → trader: buys eth on uniswap, every pool compared, only the best route runs → earn: lends idle usdg and takes it back. it can never borrow → pay for apis: pay-as-you-go data and ai, capped per api, with what you spent this week → research agent: mcp tools and paid apis on one budget → use any app: pick any contract, tick what the agent may do, everything else stays blocked → agent team: hand other agents a slice of the budget for everyday → allowance: give someone a set amount to spend. no gas, no seed phrase, take it back any time → split with friends: log who paid what, fides works out who owes who, settle your share in one go → pay many at once: paste who gets what, one bundle, one approve mix and match any of it with add-ons. a trader that also earns on what it isn't using is two ticks the terminal talk to your agents in plain english. claude does the work, inside the limits you signed. every template is a mode at the top, and if you don't have an agent for something it sets one up for you to sign. runs on your own anthropic api key, nothing goes through us every agent gets its own page: budget left, time left, what it can do in plain english, and a big stop button same contracts underneath. the fides account checks every action, and nothing moves money without your approve live now on robinhood chain → fidesbuilder.xyz/app