Vision

Decide what an agent may reach and what it may pay for.

AI agents have moved from answering questions to taking actions: calling APIs, using tools, and now paying for services on their own. Blocknet is the control point between them and the internet.

The problem

The x402 standard lets any web service ask for a payment, and lets an agent pay on the spot with a stablecoin. That opens a gap. An agent with a wallet can pay, but it can't reliably tell whether a service is trustworthy, whether it's worth paying for, or whether its last purchase from that service worked.

Ordinary firewalls and proxies act on hosts and content. They don't understand prices, budgets, approvals or the reputation of a paid service. Wallets understand money, but not policy or reputation.

Our research names four failures:

Unreliable services. Agents pay and receive errors, or nothing useful.
No refunds. Payments on-chain are final.
Weak spending controls. One runaway loop can drain a wallet a cent at a time.
Weak trust signals. Nothing records which services actually deliver.

What Blocknet does

Layer	Question it answers
Network policy	May this agent talk to this host at all?
Payment policy	Is this price within the cap, the daily budget, and the allowed networks and currencies?
Trust	Do we have reason to believe this service delivers?
Human approval	Is this unusual enough that a person should decide?
Audit	What happened, and why was it allowed?

Blocknet also learns: every paid call's outcome goes into its record, so a service that takes payment and fails is trusted less the next time.

Goals

01
Safe by default
Unknown services wait for a person. Budgets are hard limits. Blocked hosts are never contacted.
02
Non-custodial
Blocknet never holds keys, signs payments or moves funds.
03
Explainable
Every decision comes with plain-language reasons the agent can act on and a person can audit.
04
Easy to adopt
One endpoint to route calls through. Works with any agent that can make a web request.
05
Small and dependable
No third-party code, readable parts, decision logic you can test.
06
A sound business
Paid for as security and governance, not as a cut of a payments market that's still small.

What it isn't

Not a wallet or payment processor. Blocknet never takes custody. Wallet integration means the wallet asks Blocknet, not that Blocknet holds the key.
Not a content filter. It decides on hosts, prices, budgets and reputation, not on what a page says.
Not a replacement for network security. It's a policy layer. Hard enforcement needs network controls around the agent.
Not a toll on transactions. No fee on what your agents pay, for now.

Principles

Deny is cheaper than refund

Stop a bad payment before it's signed. There's no way to undo it after.

The strictest rule wins

Each check can move a decision from allow, to review, to deny. Nothing moves it back.

Policy is data, logic is pure

Your settings describe intent. The decision logic has no side effects, so it's easy to test and to reason about.

Every decision leaves a record

If it isn't in the audit log, it didn't happen.

Honest about limits

We say plainly what Blocknet doesn't enforce yet. See the limits.

Who it's for

Who	Need	Product
Teams running agents that pay for APIs	Spending limits, approvals, audit	Blocknet Gateway
Developers building agents	Control which tools and sites agents reach	Blocknet Agent

What success looks like

Near term

A team routes a real agent through the Gateway and stops paying broken or overpriced services.

Medium term

Adding Blocknet to an agent is one command, and teams run it by default, the way they run a firewall.

Long term

Wallets refuse to sign an agent's payment without a Blocknet approval, so policy is enforced where the money moves.

Next: how it works