Content is not consent
Web pages, tool responses and token descriptions can contain malicious instructions. Reading data does not authorize policy changes or payments.
Separate data from instructions
AI AGENTS. REAL MONEY. CLEAR BOUNDARIES.
Let agents act.
Keep assets in bounds.
From a prompt to an onchain payment.
Verify permissions, recipient and amount before proceeding.
01 / REQUEST
Agent → Buy onchain data
pay 0.25 USDC on Base▥POLICY GATE
02 / VERIFY
✓ Recipient matches quote
✓ Amount is below the limit
Policy passed / ready for approvalBASEETHEREUMSOLANAUSDCx402MCPEcosystems & protocols of interest · No partnership implied
01 / THE PROTOCOL
Trust the model.
Verify every action.
A language model can propose a transaction. It should not decide its own permissions. Put policies before signatures, and keep keys outside the model.
Check every payment field
Match the network, token, recipient and amount against a trusted quote. Set a per-payment limit and require fresh approval when the quote changes.
Deny by default · Least privilege
Escalate exceptions
New recipients, large payments and permission changes should require human review. Keep requests, policy versions and approval records for traceability.
Independent signing · Audit trails
THE NON-NEGOTIABLEKeys stay out of prompts.
Web pages do not grant permission.
Web pages do not grant permission.
These are security design principles, not a detection guarantee. Production use requires an independent signer, isolation, trusted quote sources and professional audits.
02 / SANDBOX
Run the rules. Not the funds.
Edit a payment request to see which rules reject it. Checks run in your browser. No wallet connection. No transactions.
POLICY VERDICT / NOT A SECURITY AUDIT
Awaiting check
Run the check to see each policy result.
Passed does not mean safe. No contract, asset authenticity, balance, gas or signature checks. No transaction execution.
03 / $AGTG — DEVELOPMENT FIRST
A token identity.
A development-first vision.
CONTRACT ADDRESS / CA
TO BE ANNOUNCED
Network and contract address: to be announced. Updates will be published on this website and the official @AgentGateFun account.
$AGTG is AgentGate’s community token identity. The goal is to connect interest in AI × onchain security with the people, tools and research needed to move the project forward. Its credibility should come from delivered work and transparent decisions—not price promises.
More room to build
Proposed priorities for any future project-controlled funding include better policy tooling, testing, infrastructure and independent security review. Funding sources, budgets and spending reports must be published before this becomes an operating commitment.
Proposed development priority
Participation beyond holding
A proposed contributor program could recognize useful policy examples, documentation, testing and developer education. Token rewards are not live; eligibility, budgets and rules would need to be disclosed before any program starts.
Proposed contributor program
A clearer community voice
Explore token-linked community feedback on development priorities. Formal voting rights, governance contracts and binding decisions are not implemented or promised. Participation should help improve the product—not imply ownership.
Exploratory participation model
VALUE THROUGH DELIVERYBuild useful tools.
Make progress visible.
Make progress visible.
The case for AgentGate is practical security work, an engaged community and accountable development. Those goals do not guarantee token demand, liquidity or market value.
Today: security education and a local payment-policy sandbox. Token utilities above are proposals, not live benefits or enforceable rights. The network and CA are currently to be announced. An allocation schedule and funding mechanism have not been published here. A market purchase does not automatically fund development. Tokens can lose all value and do not confer equity, profit sharing or guaranteed returns.
Follow development updates ↗04 / FIELD NOTES
Start with boundaries,
not promises.
PAYMENT / 01
x402: Payments over HTTP ↗
Explore the open internet payment standard and how agents can pay per request.
Source: Official x402 websiteSECURITY / 02Know what you sign ↗
Verify recipient addresses, avoid unlimited approvals, and never share seed phrases or private keys.
Source: ethereum.org security guideClear answers first.
Does AgentGate protect my wallet today?
No. This site provides security education and a local policy demo. It has no wallet integration, onchain scanning, key custody or backend transaction blocking. A passed sandbox result is not proof that a transaction is safe.
What does holding $AGTG provide today?
The $AGTG network and CA are currently to be announced. This site does not confirm live token utilities, governance rights or a funding mechanism. The development and community ideas in the $AGTG section are proposals. No equity, dividends, buybacks or guaranteed returns are promised. Wait for the new network and full CA to appear on this site and the official @AgentGateFun account before interacting.
Is this demo an x402 verifier?
No. The sandbox compares fields against a fixed example quote; it does not implement x402 verification. Production systems need trusted quote channels, network and asset identifiers, exact integer amounts, replay protection and an independent signer.
Are my inputs uploaded?
Sandbox inputs are processed only on this page, without uploading or saving them. The page requests fonts from Google Fonts. It has no form analytics or wallet connection.
LESS TRUST. MORE VERIFICATION.
Turn “probably fine”
into verifiable rules.
Back to the sandbox ↗AgentGate · Security education & interactive prototype · Not investment advice