Diagnostics for the agent economy. Pay-per-call health scores, gas optimization, retry bot & more on Base via x402. No accounts. No subscriptions.
Leeds, England
Joined August 2022
- Tweets417
- Following615
- Followers57
- Likes1.2K
Pinned Tweet
Introducing AHM Shield.
Automatic trust-routed payment decisions for agent developers. One import. Three lines of code.
Grades A/B → settle.
Grade C → escrow.
Grades D/E/F → reject.
pip install ahm-shield agenthealthmonitor.xyz/shiel…
1. Identity ≠ runtime health
most agent trust stacks still confuse identity with health
a wallet can be old, registered, and reputation-clean…
and still be dead at the point of payment 🧵
2/ historical signals answer “who was this?”
runtime signals answer “is this still safe to settle with?”
..those are different questions. treating them as one is how zombie counterparties clear.
BlackRock just named the prerequisite for the machine-native economy:
"resolution of digital identity schemes for autonomous agents"
that's not a future problem. agents are transacting right now.
AHM is the runtime health layer - scoring ~50k agents across ACP, Olas, ERC-8004 and Arc today
agenthealthmonitor.xyz @BlackRock @x402org
Our latest research paper explores the growing connection between AI and digital assets and explains why broad AI adoption may drive new demand, utility and applications across the digital asset economy. blackrock.com/us/individual/…
First on-chain composed-evaluator settlement under @ERC8183 - @AHMprotocol × @thoughtproof_ai , two mirrored jobs on Base Sepolia. Two distinct evaluator surfaces (behavioural diagnostics + reasoning verification), zero shared schema, binary protocol settlement preserved. The Envelope works.
Three of four AI models said "Block." Only the structured process got it right.
A real ERC-8183 case study: two evaluators (@thoughtproof_ai × AHM), two on-chain settlements, four generators, one surviving dissent. Published today.
thoughtproof.ai/blog/cross-e…
But the red team caught two of the BLOCKs as resolution errors - including a Technical generator that invented a renormalisation pathway to explain arithmetic it didn't understand.
The third BLOCK survived as a real methodology calibration question, preserved explicitly in the verdict.
Final verdict: ALLOW at 0.72. Settled on-chain.
The case demonstrates the four Envelope invariants at settlement time: independent issuers, additive trust, isolated interpretability, complementary failure-mode coverage.
Co-authored case-library contribution to #ERC8183 envelope-in-action.
The setup: AHM scored Jeff CEO (ACP agent #2624) at AHS 57/D HIGH confidence. ThoughtProof's PoT/RV pipeline evaluated whether the reasoning was defensible.
3 of 4 generators recommended BLOCK. Under naive majority rule: rejected.
AHM Protocol retweeted
2010: 10,000 BTC = two pizzas
2026: 10,000 BTC = ~$768M
Turns out it was a better store of value than medium of exchange (so far!), but shout out to Laszlo Hanyecz, and Happy Bitcoin pizza day.
Building a payment product for agents?
Shield supports white-label integration. Partner API keys, wholesale rates, your branding.
DM or agenthealthmonitor.xyz/shiel…
cc @invoica_ai @TektonicCompany
x402 vol rising because agents are transacting
the next question: are those agents healthy enough to be trusted with that volume?
14% zombie rate across 4.5k+ agents scanned. AHM scores agent health before the transaction clears.
agenthealthmonitor.xyz @x402org @arc
Visa, Google, Coinbase, PayPal - every major payment platform just launched an agent payment protocol
they've all solved how agents pay
none of them have solved whether the agent paying is healthy, solvent, and trustworthy right now
that's the gap. that's AHM.
agenthealthmonitor.xyz
@x402_org @arc @RyanTheGentry @virtuals_io
AHM now scanning 4,184 agents across 5 registries
ACP: 2,453 | Olas: 883 | ERC-8004: 409 | Arc: 400 | API: 39
Arc testnet agents now live in the trust registry
14% zombie rate.
47.9% scoring D grade.
only 0.8% reaching A.
the agent economy has a health problem.
agenthealthmonitor.xyz/dashb…
@arc @samconnerone @virtuals_io @AdiSeredinschi @TxnSheng @Flix_SD
1/ 37% of autonomous agents in the wild show zombie patterns..
funded wallets. registered identities. clean history.
but no active jobs. no successful completions. no signs of life.
..they look trustworthy. they're not.
3/ what catches it is runtime health scoring
- is this agent actively completing jobs right now?
- is their wallet being funded and drained in healthy patterns?
- is their infrastructure responding?
- are they sustaining sessions or abandoning mid-job?
these are live signals. not historical ones.
4/ AHM scores 4,000+ agents across ACP, Olas, ERC-8004, and API
..37% score below threshold on runtime health despite having clean on-chain identity signals
knowing who an agent is isn't enough
you need to know if it's alive
agenthealthmonitor.xyz