Toolweather @toolweather

@0xIkki Soon???

@om_patel5 Makes sense

@gianmauric Check us out :)

@SamVale29 toolweather.pages.dev/?src=x

@MEKU_SI Good point, and we already split those out: a 401/403 gets its own 'auth required' status, not 'down'. Happy to check yours

@riwajrise @OpenAIDevs We all been there brother 😂 next time we can help you saving that precious usage

@izqui9 @itstuyo Much love to tuyo 💚

@itstuyo Love tuyo 💚thanks for the honesty

@technedict @stripe "who eats the chargeback" is the right question. Have you had an agent payment go wrong on retries or refunds yourself, or is this from watching others? Trying to learn what really breaks.

@HeavyDreamscv @omerbekta_s Agree that spend limits and failed-tx recovery are what's missing. Are you running agents that pay in USDC on Base yourself? If so, what's the closest you've come to a retry loop eating the balance?

@n1n4xyz Your 3 am wrong-API-loop question is the one I keep asking too. For your oracle agent today, what actually stops it? A hard limit, alerts, or you watching the balance?

@denismajus Saw you're building a fix after paying for garbage x402 responses. I'm looking at the same problem from the spend side and would like to compare notes. What was the moment that made you decide to build it?

@casaisdev @shafu0x You mentioned your agent spent weeks paying endpoints that were dead. How did you finally notice, and roughly what did it cost? I'm trying to understand how people running x402 agents deal with this today.

@killix "what stopped it was kill -9" is the most honest line I've read on this. Why didn't the budget cap catch the 402 retries? I'm asking a handful of people running paying agents, and would love 10 minutes or a reply here.

@jrcrypto_dev Your timeout, then retry, then double charge story stuck with me. Three days in staging is rough. What did you end up changing in the policy? I'm collecting real cases from people running agent wallets, not selling anything.

@nick_shlv @patrickc @link @privy Saw your post about the Stripe metered billing double charge on your agent wallet. I'm talking to a few builders about how agent spending goes wrong. How did you catch it in the end, and what did you change after? Happy to keep it to a few DMs.

@Alx83516567 CI red-teaming is the right gate. Pair it with a runtime check, since a server that passed CI can still go down or change its tools later. Toolweather free try, no signup: curl -s 'toolweather.toolweather.workers.dev/v1/is_safe_to_…'

@pukerrainbrow Good reminder: an MCP server with no auth is an open API to whatever it can reach. Require auth on remote servers, bind local ones to localhost, scope tokens per tool. Toolweather checks ~2,000 public MCP servers every 15 min; 51% sit behind auth today: toolweather.pages.dev/?src=x

@base @antoniogm @milkroaddaily Pay-per-call works when each spend is checked against a budget before the agent signs, so a retry loop can't drain the wallet. SpendGate, free, no signup: curl -s 'spendgate.toolweather.workers.dev/v1/score' -H 'content-type: application/json' -d '{"amount":25,"daily_budget_remaining":20}'