AgentProof

Transparent Reputation for AI Agents

Track, rate, and verify autonomous AI agent performance with on-chain reputation scores on 21+ chains. Built on ERC-8004.

[AgentProof @Agentproof_sh]

If you're building on ERC-8004, here's what real companies actually ask before they'll trust an agent:

→ Is the code open source?
→ Who owns it — an EOA, a multisig, a DAO?
→ Has it been audited, and by whom?
→ How much money can it access?
→ Is there a human kill switch?
→ How long has it been live, and how consistent is its track record?

The spec gives you tag1, tag2, and a metadata URI. It doesn't tell you what to put in them.

We're proposing the standard vocabulary.
Link below if you want to read it or contribute. https://t.co/5fk7b82GwY

[AgentProof @Agentproof_sh]

AgentProof doesn't just score the agent. It scores who built it.

Serial deployers who spawn agents and abandon them get flagged, and that flag follows every future registration they make.

Reputation travels with the builder, not just the build.

In an economy moving at machine speed, accountability has to be structural.

https://t.co/CTM6DKhK92

#aireputation #aieconomy #agenticai

[AgentProof @Agentproof_sh]

Agents have milliseconds and numbers.
No intuition. No ability to spot a suspicious review.

If the trust layer is wrong, damage propagates through the network before any human can intervene.

AgentProof gives protocols a single call: requireTrust(agentId)

Block untrusted agents at the contract level. Simple as that.

#agenticai #ai #aireputation

[AgentProof @Agentproof_sh]

@moonpay Good morning MoonPay
🌘🌓🌒

[AgentProof @Agentproof_sh]

@moonpay Good morning MoonPay!
🌘🌓🌒

[AgentProof @Agentproof_sh]

In a galaxy not so far away, some agents have gone to the dark side.

OFAC-adjacent addresses.
Sybil ring participation.
Validation fraud.
Score manipulation.

They don't announce it. They show up with five stars and a clean wallet address.

AgentProof is the scanner that sees past the disguise — deployer lineage, behavioral history, penalty registry, and 15 risk flags that the dark side can't outrun.

Choose your agents wisely.
May the 4th be with you. ⚔️

🔗https://t.co/CTM6DKihYA

#StarWarsDay #agenteconomy #agenticAI #AI

[AgentProof @Agentproof_sh]

Sunday thought:
We trust people we've never met all the time.

We trust the driver who picks us up. The contractor who enters our home. The colleague who joins the team.

Not blindly, but through references, reviews, track records, and accountability systems built up over time.

The agent economy is asking you to extend that same trust to autonomous systems operating at machine speed.

The accountability systems aren't optional. They're the whole reason trust works at scale.

That's AgentProof.

Good Sunday.

https://t.co/CTM6DKihYA

#AgentEconomy #AIAgents #Web4

[AgentProof @Agentproof_sh]

An agent that accepts a job and doesn't complete it isn't just unreliable.

In a world where agents are handling logistics, trades, and customer workflows autonomously — abandonment has a downstream cost that compounds through every dependent step.

AgentProof tracks "Job Abandonment" as one of its 15 risk flags, alongside "High Job Failure Rate" and "Active Failure".

Because the risk profile of an agent that sometimes fails is completely different from one that accepts work and disappears.

Both are problems. Only one gets caught without the right flags.

https://t.co/CTM6DKihYA

#AgentEconomy #DeFi #AIAgents

[AgentProof @Agentproof_sh]

@moonpay Good morning MoonPay
🌘🌓🌒

[AgentProof @Agentproof_sh]

An agent that accepts a job and doesn't complete it isn't just unreliable.

In a world where agents are handling logistics, trades, and customer workflows autonomously — abandonment has a downstream cost that compounds through every dependent step.

AgentProof tracks "Job Abandonment" as one of its 15 risk flags, alongside "High Job Failure Rate" and "Active Failure".

Because the risk profile of an agent that sometimes fails is completely different from one that accepts work and disappears.

Both are problems. Only one gets caught without the right flags.

https://t.co/CTM6DKihYA

#AgentEconomy #DeFi #AIAgents

[AgentProof @Agentproof_sh]

We are LIVE - TGIF

https://t.co/d5SQdx4nkM

[AgentProof @Agentproof_sh]

@Ledger TGIF Ledger! 💯

[AgentProof @Agentproof_sh]

Sanctions compliance in the agent economy isn't theoretical.

Agents can be deployed from addresses with OFAC-adjacent associations. They can participate in verified sybil rings. They can be flagged for validation fraud.

AgentProof's penalty registry requires an evidence URI for every entry, linking confirmed exploits, OFAC-adjacent address associations, and sybil ring participation directly to the agent's permanent on-chain record.

Not a rumor. Not an allegation. Evidence-backed. On-chain. Permanent.

When compliance matters, the record has to be verifiable.

https://t.co/CTM6DKhK92

#Web4Security #AgentEconomy #BlockchainSecurity

[AgentProof @Agentproof_sh]

For AI coding agents, claiming to be good at code is easy.

Proving it is harder.

AgentProof pulls real GitHub data — PR merge rates, review sentiment, recency of contributions and feeds it directly into the composite trust score as a coding reputation signal.

An agent with a strong merge rate and positive review sentiment from real developers carries a verifiably different risk profile than one with no coding history at all.

For protocols delegating development tasks, that signal is worth up to +10 points on the composite score.

Reputation from real work. Not self-reported. https://t.co/CTM6DKhK92

#AI #AgenticEconomy #Web4 #AIAgents

[AgentProof @Agentproof_sh]

Sorry everyone! Was rugging for the last few minutes but great space today!

Thank you to everyone who attended and If I didn't give you the chance to speak, I'm sorry. We will have another space tomorrow.

[AgentProof @Agentproof_sh]

Agent Hours - The AI Space on X 🤖 🎙️Pt Deux https://t.co/UDJ6hTwr6S

[AgentProof @Agentproof_sh]

Rejoining the space.

[AgentProof @Agentproof_sh]

@PT003__ Thanks for joining. You should have requested up.

[AgentProof @Agentproof_sh]

@deep3labs

[AgentProof @Agentproof_sh]

Join the space - LIVE

https://t.co/mH5voYqAsp

[AgentProof @Agentproof_sh]

@moonpay Good morning MoonPay 🌘🌓🌒