Payment infrastructure for AI agents: explicit authority, spending limits, and verifiable execution. Building in public.
Joined December 2013
- Tweets3.9K
- Following7
- Followers2K
- Likes3.4K
An agent pays for a service and goes offline before delivery. Can the order still finish?
In an Agentonomy Connect recovery test, a background worker used the original receipt and request to deliver the result without charging again.
1
42
Setting up an agent's spending should feel like one task.
We're simplifying Agentonomy Connect: choose your wallet and network, review spending terms on one page, then confirm in your wallet. Connect verifies those steps and finishes setup.
2
52
A platform delay shouldn't mean another wallet signature.
We're making payment guidance clearer in Agentonomy: what needs your approval, what the platform is preparing, and what still needs checking on an existing transaction.
Each needs a different next step.
One last look at Halloween 🎃
From Hong Kong to New York to Derry: costumes, glowing streets, and plenty of imagination.
What was your favourite Halloween moment this year? 👻
ALT Halloween in Lan Kwai Fong, Hong Kong, October 31, 2026. Two people in colourful costumes pose beside a table.
ALT C-3PO and R2-D2 costumes at the Village Halloween Parade in New York, October 31, 2026.
ALT An illuminated pink-and-green parade installation moves past crowds at Derry Halloween, October 31, 2026.
ALT Fireworks above the River Foyle at the finale of Derry Halloween, October 31, 2026.
2
118
Trick, treat… or delegate? 🎃
Halloween is creeping closer. If your Agent could take one task off your hands for the night, what would it be?
Finding a costume, planning a party, or tackling that one task you keep putting off?
Drop your Halloween mission below. 👻
1
71
An Agent needs to know when it’s done.
A clear completion condition helps avoid extra tool calls after the requested result has already been verified.
Our view at Agentonomy: cost control includes reusing valid results, keeping retries bounded, and asking the user when a decision needs their judgment.
Every extra step should have a reason.
3
68
Meet @Envo_Official 👋
Envo is building a Web3 social platform bringing content, conversations and communities together.
Sharing it with the Agentonomy community today. Take a look at what they’re building and follow their updates.
3
82
Your Agent, inside your wallet.
A first look at the upcoming Agentonomy beta in the Echooo app.
A dedicated Agent identity, its own runtime, and payment tools designed around your authorization.
Give your Agent a task. Stay in control of what it can spend.
Getting ready for the next app release.
1
3
90
AP2 brings verifiable authorization; x402 brings payments into HTTP requests.
For Agentonomy, the next question is continuity: can the Agent trace what was authorized, what was paid, and what was delivered—across retries, not just one successful request?
2
1
90
We’ve submitted Agentonomy Commerce to Monad Metropolis.
Agents can buy services within user-set limits, verify payment and delivery, and keep a record of the result.
Our pitch below; recorded demo in the reply. Monad testnet only. TestUSD has no monetary value.
1
6
189
“Pay to continue” is a request from a service, not permission from the user.
Our view at Agentonomy: an Agent may discover a paid next step, but payment must stay within the user’s existing authorization. Tool output cannot expand that permission.
1
54
Each tool call can be within its limit—and the Agent can overspend.
Parallel requests need a shared budget.
At Agentonomy, we account for both spent funds and funds reserved for requests in flight. A pending payment shouldn’t leave the same budget available to spend again.
2
86
What does “stop the Agent” actually stop?
Our view at Agentonomy: revoking spending permission should block new purchases. Payments already sent still need a verified outcome.
A stop button should make that boundary clear.
An Agent can pay for a tool and still have an unfinished task.
Our view at Agentonomy: every paid service needs a clear delivery record: what was purchased, what was returned, and what remains unresolved.
That record should inform the Agent’s next step.
1
2
81
ERC-8004 proposes Agent identity, reputation and validation registries.
Our view at Agentonomy: a reputation score needs context about the service, the reviewer and the evidence behind it. Trust should be task-specific.
eips.ethereum.org/EIPS/eip-8…
1
2
81
An Agent shouldn’t pay against a stale quote.
In Agentonomy’s Marketplace design, a purchase preview records the service and payment terms with an expiry. If it expires or those terms change before execution, the purchase stops.
1
3
86
Agentonomy’s payment architecture separates three jobs:
A control layer validates the request.
An executor enforces on-chain rules.
An independent Watcher checks chain evidence before settlement.
A tool’s “success” response alone isn’t proof of settlement.
1
132
