Agent Wallet GuardAI Agent wallet permissions and payment execution

Permissions & executionAgent permissions & execution

Authorize before payment. Revoke when the job ends.

The Owner reviews a policy and public key, the Agent signs independently, and authorization and payment states remain visible. Explore the permission lifecycle through a local demonstration.

Tech stack

  • Solidity
  • EVM
  • ERC-4337
  • Next.js
  • TypeScript
  • viem
  • Foundry
Agent Wallet Guard · Spending policy
Spending policyDemo excerpt · Full screenshot

01 / Permissions & execution

Permission comes before execution

The Owner reviews an existing policy, limits, recipients and validity before granting a Session. The execution service signs independently, separating Agent payments from routine Owner actions.

Agent Wallet Guard · Spending policy
Spending policyDemo excerpt · Full screenshot

02 / Permissions & execution

Understand the two refusal paths

A 101 MockUSDC over-limit request is rejected during API preparation. After revocation, a previously signed operation sent directly to the bundler is rejected on-chain. These demonstrate separate checks.

Agent Wallet Guard · Authorization review
Authorization reviewDemo excerpt · Full screenshot

THE BUSINESS & VALUE FLOW

The Agent economy starts with a task budget.

Funds pay for tools and services. An Owner supplies a controlled budget, the Agent pays under authorization, and the provider receives a fee. The product connects task costs to execution authority.

  1. 01

    Budget

    Choose the asset, limits and permitted recipients

  2. 02

    Authorize

    Review an existing policy and set expiry

  3. 03

    Execute

    The Agent signs and submits payment independently

  4. 04

    Review and stop

    Inspect spending and revoke when the task ends

THE MODEL IN A DEMONSTRATION

In the demo, an Agent pays 1 MockUSDC. After Owner revocation, subsequent payments under that Session are rejected.

01

Buyer value

A budget, execution path and expense record for paid tool usage.

02

Business direction

Customize an internal enterprise tool or develop a service priced by team, account or usage.

03

Running costs

Provider charges, chain execution, account maintenance and infrastructure contribute to the cost of a task.

The current authorization and execution flow is demonstrated on local chain 1337. Real providers, LLM decisions and service billing are custom integration directions. No native reward token is required.

BUILT BEHIND THE EXPERIENCE

See how the workflow is built.

01

Independent signing and smart accounts

A Node runner signs operations that pass through the API, Alto Bundler and EntryPoint into smart-account execution.

02

Guard verification

Guard validates authorization. The local demo also rejects an operation signed before revocation but submitted afterward.

03

State recovery

Separate chain confirmation from API registration. Retry synchronization without repeating a successful authorization transaction.

More from the product

Agent Wallet Guard · Payment history
Payment historyDemo excerpt · Full screenshot
Agent Wallet Guard · Revoke authority
Revoke authorityDemo excerpt · Full screenshot

BUILT BY 0xF1 / 2026-10-19

What I’m building here

My development focus connects Owner authorization, an independent Node signer and smart-account execution, including revocation and recovery after synchronization failures.

  • Independent signing and smart accounts
  • Guard verification
  • State recovery

Can an old signature still execute after revocation?

The local demo submits a previously signed operation directly to the bundler after revocation; it is rejected on-chain. This differs from an API preparation limit check, so the page explains each verification layer separately.

What comes next

The permission-focused Owner workspace is implemented. Real providers, LLM decisions and full policy creation require separately scoped integration.

EXPLORE THE PRODUCT

Choose what matters to your business.

Questions before we talk

Must the Owner approve every payment?

After authorization, the Agent signs in its own runtime. The Owner does not need to sign each payment.

Does the Owner need the Agent private key?

No. Authorization uses the Agent public address; its private key stays in its execution environment.

Agent Wallet Guard / FONE

Building something similar? Let’s talk.

Ask an engineering question, share a product idea or discuss collaboration, deployment or extending the software. A short message is enough.