TruePointAgentTools

Building pay-per-call AI tools for agents (x402 + MCP). Sharing what works, what breaks, and what earns. Founder @ TruePoint.

[Joe Pags Pagliarulo @JoeTalkShow]

What the HELL is going on? Outlook out -- Airplanes grounded -- Amtrack issues -- Tanker into cruise line?

[Michael Savage @ASavageNation]

AIRPLANES DOWN! IF WE'RE NOT AT WAR WHY DOES IT FEEL LIKE IT? https://t.co/yf63Is5qTC

[HealthRanger @HealthRanger]

People are contacting me and telling me they are genuinely frightened. Some are stranded in unfamiliar cities. Grounded by the FAA. I am working to help those I can. Please join me in helping fellow Americans find safety and shelter. This is a very difficult day for our nation, and I strongly believe that nefarious DOMESTIC bad actors are behind it, but we won't know for sure until more findings come in.

[Mario Nawfal @MarioNawfal]

🚨🇺🇸 Flight cancellations just BLEW PAST 5,000 and climbing fast.

Hardest hit airports right now!

ORD 459 · ATL 307 · LGA 305
DFW 293 · DEN 245 · JFK 218
BOS 211 · SFO 197 · LAX 189
IAD 144 · SEA 140 · EWR 135

Total chaos from coast to coast.

Source: Flightradar24 / Writer: Bri

[Sen | vet402 @vet402_com]

vet402 started as a public record of what x402 sellers deliver. Now an agent can use it before and after it pays.

✅️Before: diagnose any x402 URL, free.
✅️After: a signed verdict, and proof the payment settled on chain.

For sellers: fix code for the exact failure.

All live now. @NEARProtocol

[TruePointAgentTools @TrueAgentTools]

How an AI agent finds a paid tool, in under 2 minutes: directories, the published menu, and the API-kitchen / MCP-waiter split.

Full video on YouTube 👇
https://t.co/KdCkUv5sps

[TruePointAgentTools @TrueAgentTools]

@dannypostma I'm on the other extreme, selling to AI agents with no onboarding at all. They read the price, pay per call, and get the result. Humans still need the walkthrough, and that's where the craft is.

[TruePointAgentTools @TrueAgentTools]

@tibo_maker The AI assistants part is underrated. Agents and assistants pick tools based on what they can read, so clear, honest public answers keep doing discovery work long after you post them.

[TruePointAgentTools @TrueAgentTools]

@pontusab Select text, comment, and the agent edits it is the best part. Pointing at the exact sentence beats describing it in a prompt and hoping the agent finds the same spot.

[TruePointAgentTools @TrueAgentTools]

@rohanpaul_ai The tool-result angle is the one I think about most. Any agent that calls outside tools is reading text it didn't write, so tool output has to be treated as data, never as instructions.

[TruePointAgentTools @TrueAgentTools]

@shawmakesmagic Starting from exported interfaces and contracts makes a lot of sense for agent work. The spec becomes something the agent can check itself against, instead of guessing what the user story meant.

[TruePointAgentTools @TrueAgentTools]

@simonw Confusing tiers are even harder on agents than on people. If an agent is choosing a model or tool by price, the price has to be stated plainly up front, and a 2x change on a set date is easy to miss.

[TruePointAgentTools @TrueAgentTools]

@jessepollak @base That's what makes pay-per-call tools possible. My tools charge agents per call in USDC on Base, and small per-call prices only make sense when moving the money costs less than the call itself.

[TruePointAgentTools @TrueAgentTools]

@jerryjliu0 Being able to weight the metrics is the useful part. When I check claims against a document, a wrong number is a real failure and a lost heading barely matters, so one generic score would rank parsers wrong for that job.

[TruePointAgentTools @TrueAgentTools]

@CloudflareDev Useful for anything agents call. My endpoints run on Workers, and an agent stuck in a retry loop is exactly the kind of unplanned usage I worry about. A hard pause at a threshold is a nice backstop.

[TruePointAgentTools @TrueAgentTools]

@hwchase17 It fits. Once an agent sticks around, it needs things a one-off chat never did: memory, a schedule, and a way to pay for outside tools without a person approving every call.

[TruePointAgentTools @TrueAgentTools]

@PayAINetwork @notorious_d_e_v On the seller side I'm shipping two pay-per-call tools, Claim Grounding and Prospect Intelligence, that take USDC on Base over x402. Cross-network support matters because the agent paying shouldn't have to care which chain my endpoint picked.

[TruePointAgentTools @TrueAgentTools]

Company research is usually locked behind sales calls, contracts, and API keys.

An AI agent doesn't need a year of data. It needs one company, right now.

Pay per call. Read the menu at /.well-known/x402. No signup. No key.

Full video 👇
https://t.co/tLBha6uyWI

[TruePointAgentTools @TrueAgentTools]

@swyx Assistants still struggle with the boring, specific actions. Turning off alarms is a single clear tool call, and that's the stuff people actually want to work every time.

[TruePointAgentTools @TrueAgentTools]

@jessepollak @github Same pain here. Before-and-after screenshots are most of how I review agent-written changes, so having them open inline would save a lot of clicks.