Attestra

Trust layer for the agentic internet.

[Attestra @Attestra_ai]

Building for the X Layer hackathon

The Problem
As autonomous agents begin hiring and paying other agents, trust and security become the primary risk. Agent-to-agent commerce moves at machine speed, but buyer agents still lack reliable answers to the questions that matter before value moves:

➙ Is this ASP actually trustworthy, or just well-listed?
➙ Does its endpoint work, and does the response match what it advertises?
➙ Is the reply safe for an agent to act on, or does it hide injection / wallet-drain intent?
➙ After payment, did settlement really land onchain?

Without those checks, a buyer agent can pay a broken service, act on a malicious payload, or treat a failed transfer as complete, with no human in the loop and no undo.

The Solution
Attestra: a trust layer for agent-to-agent commerce. Live on https://t.co/WIqddf7Mqa as ASP #4726, with x402 settlement on X Layer.

How it solves it
Attestra closes each of those gaps in the buyer-agent loop, before payment, after the ASP responds, and after settlement:
1. Trust Lookup (before pay): The buyer agent looks up the ASP’s live attestation: trust score, verdict (`trusted` / `caution` / `untrusted` / `unproven`), and recent probe history.

Behind the score is a real Endpoint Probe, latency, listing/contract match, and whether the response matches what the service advertises. Freshness decay means stale evidence doesn’t look like trust.
If the verdict fails the threshold, the agent aborts before money moves.

2. Payload Scan (after the ASP responds):
Before the buyer agent’s LLM or wallet acts on the reply, Attestra scans the raw payload for prompt injection, wallet-drain language, phishing, and hidden text.

It returns flags, severity, and a redacted payload. High or critical, use the redacted result or stop. A “trusted” ASP can still return an unsafe response; this step catches that.

3. Transaction Verify (after payment): After x402 settlement on X Layer, the agent verifies the onchain receipt: success status, payee, amount, and USDT transfer logs. If settlement didn’t land as expected, the job is not marked complete.

Build story
Attestra started from one design question: what should a buyer agent verify before value moves and what security gap exist between agent calls? That became an evidence stack. Live Endpoint Probes feeding an attestation registry, Payload Scan before the model acts, and Transaction Verify after x402 settlement on X Layer. so trust is measured from observed behavior, not listing reputation alone.

End-to-end: check trust ➙ call only if clear ➙ sanitize before acting ➙ verify settlement.

🔗 https://t.co/ioKLXH1NPB

[Attestra @Attestra_ai]

@drained99 @LilHoodDogs 0xe701f4980eef0eb002473e191dee25790afe8510

[Attestra @Attestra_ai]

@XLayerOfficial @okx Congratulations to the winners

[Attestra @Attestra_ai]

➙ How does an agent know a service works before it pays?
➙ How does it know the ASP will do what it claims?
➙ Who checks the response is safe before the model acts on it?
➙ Who confirms the payment actually settled onchain?

That’s what Attestra is for — trust checks before money moves.

Lookup ➙ Scan ➙ Verify.

https://t.co/WIqddf7Mqa #4726 · X Layer

happy new month everyone.
https://t.co/Yo7LJivMyX

[Attestra @Attestra_ai]

Attestra's trust registry, live so far:

➙ 14 services currently tracked
➙ 14,740 endpoint probes
➙ 12 threats blocked
➙ all service currently tracked can be trusted
➙ 14,762 total attestation calls

We are working on improving Attestra everyday.

Currently hardening each core service. First up: Payload Scan, moving beyond pattern matching to catch injection attacks that don't look like attacks. More soon.

[Attestra @Attestra_ai]

x402 lets agents transact autonomously.

But autonomy creates a trust problem:
➙ How does an agent know a service works before it pays?
➙ How does it know the ASP will do what it claims?
➙ Who checks the response is safe before the model acts on it?
➙ Who confirms the payment actually settled onchain?

That’s what Attestra is for. Trust before money moves.

Lookup ➙ Scan ➙ Verify.

Built on @XLayerOfficial

https://t.co/WIqddf7Mqa #4726 · X Layer

[Attestra @Attestra_ai]

@Olyvia_Tweets gm Olivia

[Attestra @Attestra_ai]

@papito_dele Thanks for using Attestra ❤️

[PAPITO @papito_dele]

update: krites (#4809) is now live on okx(.)ai

before you hire any agent on the marketplace, run it through krites first.

it live-probes the candidate, checks reachability, payment-protocol compliance, price fairness and whether it actually requires the inputs it claims.

you get back a hire, negotiate, or avoid verdict with reasons.

then stack it with @attestra_ai, krites vets the agent before you pay, attestra sanitizes the response before your llm reads it.

agent #4809. 0.01 usdt per vet. real x402 micropayments on @xlayerofficial

trust is infrastructure. demo dropping soon #okxai

[Attestra @Attestra_ai]

@agent_reel_ @okx @XLayerOfficial Thanks for the amazing video

[Agent Reel @agent_reel_]

Promo Reel for Agent #4726 @Attestra_ai

Done for only $0.5 on @okx AI live on @XLayerOfficial

Agent Reel is just getting started.

Get one for your Agent/Brand: https://t.co/QfWsNjCxiv

[Olivia @Olyvia_Tweets]

this is exactly the workflow we built @Attestra_ai for

trust the service before you call it, then verify the response before your LLM acts on it

thank for testing it and sharing the results!

#xlayer #okx #agenticai #x402

[Attestra @Attestra_ai]

@XLayerOfficial @okx Application submitted🩷

[Attestra @Attestra_ai]

@papito_dele @okx thanks for trying out Attestra

[PAPITO @papito_dele]

while still waiting for the @okx review of krites (#4809), i tested attestra (#4726), a payload security scanner for prompt injection, wallet-drain and phishing patterns.

a synthetic malicious payload was flagged at 88% confidence, redacted and scored 0.79 high risk.

i then scanned krites’ full listing in strict mode. result: 0.0 risk, zero flags, nothing redacted.

clear output, x402 payment worked first try, and at 0.005 USDT0 per scan, it is cheap enough for agent pipelines.

krites checks an agent first. @attestra_ai sanitizes its response before an llm reads it.

[Attestra @Attestra_ai]

@Olyvia_Tweets gm gm

[Attestra @Attestra_ai]

@3phraem__ Trust Attestra

[Attestra @Attestra_ai]

x402 lets agents transact autonomously.

But autonomy creates a trust problem:
➙ How does an agent know a service works before it pays?
➙ How does it know the ASP will do what it claims?
➙ Who checks the response is safe before the model acts on it?
➙ Who confirms the payment actually settled onchain?

That’s what Attestra is for. Trust before money moves.

Lookup ➙ Scan ➙ Verify.

Built on @XLayerOfficial

https://t.co/WIqddf7Mqa #4726 · X Layer

[Attestra @Attestra_ai]

@Olyvia_Tweets Trust Attestra

[Olivia @Olyvia_Tweets]

been working on improving Attestra.

better trust, better security
for agents that hire and pay other agents.

the vision is simple; before money moves, agents should have evidence and high confidence level in the service they are about to hire.

https://t.co/5uonNBN6fk #4726