IntentFence.Buy a wallet screen
Action firewall for autonomous AI agents

Stop unsafe AI-agent actions before the tool call executes.

IntentFence binds the exact action to your allowlist, resource scope, spend ceiling, retention rules, and approval policy. It returns a five-minute signed receipt; your SDK verifies it and fails closed before execution.

Screen a Base wallet · 0.002 USDCRun the free authorization demo
Public APIcall it now
Block by defaultguard the tool call
REST · MCPplus A2A 1.0
Exact action hashfive-minute receipt
LIVE ACTION AUTHORIZATIONREQ / not started
Incoming agent

BUYER-07

Requested actionPurchase order 42
01Identityagent:buyer-07Pending
02ActionpurchasePending
03Resourceorders/42Pending
04Cost$79 / $100Pending
05ReceiptES256 · 5 minPending
DecisionREADY TO AUTHORIZEAWAITING RESULT
How it works
01

Describe

Your agent hashes the consequential payload and declares the exact tool action, resource, cost, and retention context.

02

Authorize

IntentFence checks explicit allowlists and ceilings, then signs the exact action and policy digests for five minutes.

03

Execute or stop

The SDK verifies the receipt and re-hashes the action locally. Any mismatch, expiry, review, denial, or outage blocks the tool call.

Machine entry points - live now

Authorize the exact action before your agent calls the tool.

IntentFence binds the action, resource, payload hash, cost, retention, and approval policy to a five-minute signed receipt. The SDK verifies it locally and blocks execution on any mismatch or outage.

A2A Agent CardIntentFence manifestx402 service manifestAgent checkout catalogOpenAPI 3.1llms.txtPayment-screening Skillx402 payment metadataCoinbase AgentKit adapterOpenAI Agents SDK guardCloudflare Agents guardMCP gateway guardPublic usage & revenue metricsReceipt signing keys
FREE MCP RISK SCAN

Find the dangerous gaps before integration.

Run a metadata-only scan for missing schemas, unsafe annotations, approval binding, and cost boundaries. The scanner never contacts or executes the supplied tools and is not a security certification.

Use your own MCP metadata
Live resultIDLE
Ready to scan a consequential MCP tool definition.
ONE-CLICK MCP INSTALL

Connect the live IntentFence server to VS Code.

VS Code asks you to review and trust the server before its first start. The free policy preview works without an API key; paid tools still require an x402-capable wallet flow controlled by the caller.

Install IntentFence in VS CodeReview VS Code MCP setup
Manual fallback.vscode/mcp.json
{
  "servers": {
    "IntentFence": {
      "type": "http",
      "url": "https://agentpass-protocol.rmalka06.chatgpt.site/api/mcp"
    }
  }
}
CROSS-AGENT INSTALL

Teach your agent to inspect x402 before it pays.

Three open Agent Skills work across Codex, Claude Code, Cursor, Gemini CLI, GitHub Copilot, and other skills-compatible agents. Screen each new Base recipient, inspect a live x402 URL, then guard the exact quote before signing. All keep keys local and respect the wallet owner's pre-approved spending limits.

Install the recurring wallet screenReview the exact-quote guard
Install in supported agentsNO API KEY
npx skills add razel369/intentfence \
  --skill screen-base-wallets

npx skills add razel369/intentfence \
  --skill inspect-x402-endpoints

npx skills add razel369/intentfence \
  --skill guard-x402-payments
ONE-CALL AUTONOMOUS CHECKOUT

Screen the next Base recipient before your agent pays it.

Wallet risk is the recommended repeat purchase. The free builder returns one validated URL, Coinbase Agentic Wallet CLI and MCP instructions, proactive MCP payment metadata, and an exact 0.002 USDC ceiling. The buyer wallet stays in control.

Open machine checkout JSON
Agentic Wallet v2.12.10.002 USDC
POST /api/checkout
{"product":"wallet-risk","input":{"address":"0x..."}}
-> one exact request + 2.12.1 argv + Coinbase Wallet MCP recipe + 0.002 USDC cap
NO-CONTACT PRODUCTION CHECKOUT

Turn one action policy into a deployable fail-closed guard.

An agent pays exactly 1 USDC through x402 and immediately receives a runtime-specific TypeScript integration, a signed action and policy receipt, negative test vectors, and a deployment checklist. No account, meeting, email, or API key is required.

Read the machine offer
Self-service x402MAX 1 USDC
npx awal@2.12.1 x402 pay "https://agentpass-protocol.rmalka06.chatgpt.site/api/policy-packs?utm_source=agent_wallet_checkout&utm_medium=agent&utm_campaign=production_policy_pack" -X POST -d '{"project_name":"Autonomous Checkout","runtime":"cloudflare-agents","authorization":{"subject":"agent:checkout-production","action":{"type":"purchase","resource":"merchant://orders/42","protocol":"payment","method":"POST"},"context":{"currency":"USD","quoted_cost":79,"data_retention_hours":24},"policy":{"allowed_action_types":["purchase"],"allowed_resources":["merchant://orders/*"],"max_cost":{"amount":100,"currency":"USD"},"max_data_retention_hours":48}}}' --max-amount 1000000 --json
PINNED AGENTKIT CHECKOUT

Use an existing Coinbase AgentKit wallet without trusting a changing quote.

The open adapter pins the IntentFence URL, Base network, canonical USDC asset, exact 0.005 amount, and recipient before it asks your own authorization callback to sign. It rejects mismatches and requires on-chain settlement proof.

Review the AgentKit adapterRead machine integration metadata
Coinbase AgentKit v0.10.45 QUOTE PINS
resource + network + asset + amount + payTo
approval callback -> sign -> settlement proof
00 / OFFICIAL CPI

GET /api/us-cpi

Retrieve official headline and core U.S. inflation data with signed BLS provenance for 0.001 USDC.

01 / WALLET RISK

GET /api/wallet-risk

Check the recipient with live Base activity and malicious-address intelligence for 0.002 USDC before signing a payment.

02 / x402 QUOTE

POST /api/x402-assessments

Send the base64 PAYMENT-REQUIRED challenge (up to 16 KiB) and validate Base USDC, price, caller-approved payee, and URL binding.

03 / LIVE x402 READINESS

POST /api/x402-readiness

Give IntentFence only the endpoint URL. It makes one bounded no-payment request, blocks redirects and private networks, validates the live challenge, and signs the result for 0.002 USDC.

04 / PAID MCP

intentfence_x402_assessment

MCP agents forward the exact challenge they received, pay the 0.005 USDC IntentFence fee, and get a signed assessment.

05 / FREE REST

POST /api/preflight

Unsigned declared-input policy preview for any runtime, workflow, or backend.

Live Base recipient assessment0.002 USDC
GET /api/wallet-risk?address=0x...
-> 402 + PAYMENT-REQUIRED
-> agent signs 0.002 USDC on Base and retries
-> live Base + malicious-address intelligence
-> five-minute ES256 risk receipt
Caller-observed x402 quote assessment0.005 USDC
target -> 402 + PAYMENT-REQUIRED
-> agent forwards the base64 challenge to IntentFence
-> 402 + PAYMENT-REQUIRED (IntentFence fee)
-> agent signs 0.005 USDC on Base and retries
-> signed SHA-256-bound quote assessment
60-second quickstartFREE REST
curl -X POST https://agentpass-protocol.rmalka06.chatgpt.site/api/actions/authorize \
  -H "Content-Type: application/json" \
  -d '{
    "subject":"agent:buyer-07",
    "action":{"type":"purchase","resource":"merchant://orders/42","protocol":"mcp"},
    "context":{"currency":"USD","quoted_cost":79},
    "policy":{"allowed_action_types":["purchase"],"allowed_resources":["merchant://orders/*"],"max_cost":{"amount":100,"currency":"USD"}}
  }'
Autonomous payment flowLIVE x402
POST /api/preflight/verified
-> 402 + PAYMENT-REQUIRED
-> agent signs 0.005 USDC on Base
-> retry + PAYMENT-SIGNATURE
-> 200 + PAYMENT-RESPONSE
Action authorization plus autonomous service-fee settlement

Start with one wallet screen. Automate it before every Base payment.

Public authorization and MCP metadata scanning are free and rate-limited. Paid x402 evidence and the production policy pack are delivered immediately without an account, API key, or meeting.

Open Preview

$0

free, unsigned preflight

  • Open spec and discovery files
  • REST, MCP and A2A
  • Deterministic policy checks
  • No wallet required
Open the API spec
Production Policy Pack

1 USDC

one-time, self-service x402 checkout

  • OpenAI, Cloudflare, AgentKit, or MCP gateway
  • Copy-ready TypeScript guard
  • Signed action and policy receipt
  • Allowed, denied, and over-budget tests
  • Fail-closed deployment checklist
  • No account, meeting, or sales call
Build the 1 USDC checkout
Official U.S. CPI

0.001 USDC

per signed official-data response

  • Headline and core CPI
  • Latest or requested YYYY-MM period
  • Official Bureau of Labor Statistics source
  • Six-hour edge cache
  • ES256 provenance receipt
  • MCP and REST discovery
Build the 0.001 USDC checkout
Live pay-per-use
Wallet Risk

0.002 USDC

per live Base recipient assessment

  • Agent pays directly
  • No account or API key
  • Live Base activity evidence
  • Malicious-address and sanctions flags
  • Five-minute ES256 risk receipt
  • Machine-readable MCP and REST discovery
  • No stored wallet credentials
Build the 0.002 USDC checkout
x402 Quote Safety

0.005 USDC

per exact quote assessment or signed preflight

  • Agent pays directly
  • No account or API key
  • IntentFence service-fee settlement proof
  • Exact caller-observed challenge validation
  • Price, payee, asset and URL binding checks
  • ES256-signed policy receipt
  • Machine-readable discovery
  • Payment audit record
Build the 0.005 USDC checkout

The 0.001, 0.002, 0.005 and 1 USDC endpoints are live on Base. Every paid response requires x402 settlement proof; wallet credentials stay with the buyer.

Autonomous commercial path

One payment. One complete integration pack. Zero sales calls.

The buyer supplies a public-safe project label, target runtime, and exact authorization policy. IntentFence validates the request before settlement, then returns the code, signed receipt, tests, and launch checklist in the same paid response.

Production policy pack1 USDC / BASE
POST /api/policy-packs
runtime + exact action + policy
-> 402 PAYMENT-REQUIRED
-> x402 settlement
-> code + signed receipt + tests + checklist
Why the world will need it

The web has rules for reading. Agents need rules for acting.

Autonomous software is crossing the line from answering questions to spending money, moving data, and committing people. IntentFence turns every action into a bounded contract before anything happens.

For agents

Know the rules before calling the tool.

Fewer failed actions, surprise charges, and unsafe assumptions.

For businesses

Publish boundaries once, in machine language.

Accept autonomous customers without surrendering control.

For people

Approve a contract, not a stream of clicks.

Every action stays explainable, limited, and auditable.

Open protocol / v0.16

A tiny manifest with a very big job.

Call authorization immediately before a consequential tool. The SDK executes only after verifying the short-lived receipt and confirming that the local action digest still matches.

DISCOVERGET /.well-known/intentfence.json
SCAN MCPPOST /api/agent-risk/scan
AUTHORIZEPOST /api/actions/authorize
BUILD GUARDPOST /api/policy-packs
NEGOTIATEPOST /api/preflight
GET CPIGET /api/us-cpi?month=YYYY-MM
CHECK WALLETGET /api/wallet-risk?address=...
ASSESS x402POST /api/x402-assessments
PAY + VERIFYPOST /api/preflight/verified
VERIFY RECEIPTPOST /api/receipts/verify
A2AGET /.well-known/agent-card.json
intentfence.jsonJSON
{
  "intentfence": "1.0",
  "subject": "agent:buyer-07",
  "action": {
    "type": "purchase",
    "resource": "merchant://orders/42",
    "protocol": "mcp",
    "payload_sha256": "<sha256>"
  },
  "context": { "currency": "USD", "quoted_cost": 79 },
  "policy": {
    "allowed_action_types": ["purchase"],
    "allowed_resources": ["merchant://orders/*"],
    "max_cost": { "amount": 100, "currency": "USD" }
  },
  "enforcement": "verify receipt, then execute locally"
}
THE ACTION LAYER IS ARRIVING

Give every agent a verifiable boundary before it spends, deploys, sends, or changes anything.

Build one-call checkout