Autonomous software is becoming an economic actor

Track the stack for AI agents that can hold wallets, pay with stablecoins, and work on-chain.

A practical field guide for builders, crypto founders, investors, and technical readers following the intersection of AI agents and programmable money.

Start with the guide Explore the tool stack
Agent wallets Stablecoin rails API payments On-chain identity DeFi automation

On This Page

Use the site as a publication checklist, not a recommendation list.

Beginner Guide

What is an AI crypto agent?

An AI crypto agent is software that uses model reasoning, external tools, and blockchain accounts to complete financial or operational tasks with limited human intervention. The crypto layer gives the agent a way to pay, receive funds, prove identity, and interact with protocols.

01

Wallet control

The agent needs a wallet, policy engine, key management, and transaction signing flow. Most production systems use delegated permissions rather than handing a model a raw private key.

02

Payment intent

Stablecoins make agent-to-API and agent-to-agent payments predictable. The useful question is not whether an agent can pay, but what budget, limits, and refund path govern the payment.

03

Protocol interaction

Agents can rebalance positions, source liquidity, post collateral, or execute workflows across smart contracts. Guardrails and simulation matter more than speed.

04

Reputation and auditability

On-chain identity, attestations, and transparent activity logs help counterparties decide which agents are safe to trust, rate-limit, or insure.

Educational Guides

Twelve in-depth articles on AI agents and programmable money.

Educational content only. Nothing here is financial advice. Browse the full article index or start with the pillar guide below.

Month 1 · Explainer

AI Agents and Stablecoins

How autonomous software may use USDC, USDT, and other dollar-pegged rails for micropayments.

Month 2 · Comparison

AI Agent vs. Trading Bot

Architectural differences, use-case fit, and when each pattern may apply.

Related sites: Stablecoin Payments · Crypto Education · Security Audits

Educational Model

How an agent payment should move from intent to audit trail.

This sequence is a learning model for builders. It is not a production integration guide or financial advice.

1. Intent

Record what the agent wants to do, which account it will use, why the action is allowed, and who can review it.

2. Policy check

Compare the request with budgets, allowlists, chain limits, counterparty rules, and human approval requirements.

3. Simulation

Preview balances, approvals, fees, slippage, and failure modes before any wallet signs a transaction.

4. Settlement

Submit only the approved transaction or payment proof, then capture receipts and transaction hashes.

5. Monitoring

Watch status, retries, refunds, and policy drift so a stuck or unsafe agent can be paused quickly.

6. Review

Keep prompts, tool calls, source references, and operator decisions available for later audits.

Tools Directory

The AI agent crypto stack

Showing all tool categories.

Smart wallets

Account abstraction wallets with spending limits, session keys, and recoverability.

Use for: agent-specific policy accounts

MPC custody

Distributed signing infrastructure that avoids exposing full private keys to one process.

Use for: institutional agent operations

Transaction simulation

Pre-flight checks for swaps, approvals, transfers, and contract calls before execution.

Use for: reducing irreversible mistakes

Stablecoin payment APIs

Rails for USDC, USDT, and other stable assets across L2s and high-throughput chains.

Use for: API metering and global settlement

Micropayment gateways

Pay-per-request flows for model calls, data feeds, inference, storage, or scraping tasks.

Use for: machine-to-machine commerce

On-chain invoicing

Receipts, metadata, and accounting exports tied to programmable payments.

Use for: finance operations

DIDs and attestations

Identifiers and credentials that let agents prove affiliation, authorization, and history.

Use for: trust and access control

Reputation registries

Activity-based scoring, allowlists, slashing hooks, and public agent track records.

Use for: marketplace safety

Human delegation proofs

Credentials that show which human, DAO, or company is responsible for an autonomous workflow.

Use for: compliance and support

DeFi routers

Aggregation layers that choose liquidity venues while respecting slippage and execution constraints.

Use for: swaps and rebalancing

Risk engines

Rules, scoring, and monitoring that prevent agents from taking unmanaged protocol exposure.

Use for: autonomous treasury policies

Oracle and data feeds

Price, volatility, liquidity, and state data that agents use before executing transactions.

Use for: decision grounding

Policy engines

Deterministic rules that translate human constraints into budgets, scopes, pauses, and approvals.

Use for: enforceable agent limits

Revocation controls

Operational paths to pause sessions, rotate keys, cancel permissions, and recover from unsafe behavior.

Use for: incident response

Audit trails

Append-only records of prompts, policy decisions, payment intents, signatures, and settlement results.

Use for: accountability reviews

Research Standards

How this guide treats sources and risk.

Listings are educational references, not endorsements, rankings, or investment recommendations.

Preferred sources

Use project documentation, protocol docs, technical blogs, audits, standards drafts, and public repositories.

Risk context

Describe custody, smart contract, oracle, compliance, and operational risks before discussing use cases.

Freshness

Treat integrations, fees, supported chains, and APIs as time-sensitive details that require verification.

No advice

Do not frame tokens, protocols, or infrastructure providers as buys, yield opportunities, or guaranteed outcomes.

Source review workflow

Keep profiles in placeholder status until official docs, repositories, security notes, and review dates are recorded.

Article review gate

Publish article drafts only after claims, examples, risks, and no-advice language are checked against the static templates.

Project Profiles

10 projects to watch

These profiles are editorial starting points for research. They describe categories and practical relevance, not investment advice.

Wallet infrastructure

Safe

Multisig and smart account infrastructure widely used by teams that need policy-driven treasury control.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Identity

Ethereum Attestation Service

Composable attestations that can help agents prove roles, approvals, and off-chain facts on-chain.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Stablecoins

Circle USDC

Dollar settlement rails for predictable agent payments, subscriptions, and global payout workflows.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Agent payments

Coinbase x402

A payment pattern for HTTP APIs where clients can pay resources directly as part of request flow.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

DeFi liquidity

Uniswap

Core decentralized exchange infrastructure agents can use for routing, swaps, and portfolio automation.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Price data

Chainlink

Oracle networks and automation services that can ground agent decisions in market and state data.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Storage

IPFS

Content-addressed storage useful for agent logs, public artifacts, and portable data references.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Agent frameworks

LangChain

Tool-calling and orchestration patterns that help connect LLM agents to wallet and data actions.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Crypto AI network

Fetch.ai

Agent-oriented infrastructure and marketplace concepts for autonomous services and coordination.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Compute markets

Akash Network

Decentralized compute marketplace that maps to agent workloads needing flexible infrastructure.

Source review pending: add official docs, public repository, audit notes, and latest review date before publication.

Article Briefs

Educational outlines that need source review before publication.

These drafts are content scaffolds. They should not be treated as tutorials, recommendations, or operational runbooks. Open the local article index and local decision log for review status.

Brief 01 · 8 min read · Source status: draft

Build an agent wallet with a daily stablecoin budget

Explain how teams can model a scoped wallet, spending policy, simulation step, and emergency pause without exposing private keys to a model.

  1. Create a dedicated account for the agent.
  2. Add a spending policy and an emergency pause signer.
  3. Log every payment intent before signing.
Open detail scaffold

Review gate: verify wallet policy terminology, custody assumptions, and non-production framing.

Brief 02 · 10 min read · Source status: draft

Pay an API with a stablecoin micropayment

Walk through a payment-required response, budget check, payment proof, retry, and receipt trail for a metered API request.

  1. Return a payment requirement from the API.
  2. Let the agent evaluate price and budget.
  3. Attach proof of payment to the second request.
Open detail scaffold

Review gate: confirm refund wording, network-fee caveats, and local-only example boundaries.

Brief 03 · 12 min read · Source status: draft

Run a guarded DeFi rebalance

Describe a supervised rebalance flow where an agent reads balances, compares quotes, simulates the transaction, and records why a human-approved action was allowed.

  1. Read balances and target allocation.
  2. Choose a route only after quote comparison.
  3. Execute if risk checks and simulation pass.
Open detail scaffold

Review gate: remove yield language, add slippage caveats, and keep examples asset-neutral.

Brief 04 · 9 min read · Source status: draft

Design revocation for an autonomous wallet session

Teach readers how to think about time limits, scoped permissions, cancellation paths, and incident notes when an agent session should stop acting.

  1. Define who can pause or revoke the session.
  2. List the budgets, contracts, and actions the session can reach.
  3. Document the evidence needed before restoring access.
Open detail scaffold

Review gate: verify access-control terms and avoid implying any wallet pattern is universally safe.

Brief 05 · 11 min read · Source status: draft

Create an audit trail for agent payments

Outline the records a reviewer needs after an agent makes or attempts a payment: prompt context, policy result, simulated outcome, approval, settlement proof, and post-action note.

  1. Capture intent, policy decision, and reviewer identity.
  2. Link simulation output to the final signed action.
  3. Keep failed attempts visible for later risk review.
Open detail scaffold

Review gate: check privacy language, retention assumptions, and non-advice disclaimers.

Source-Gap Issue Log

Unresolved source gaps that block publication-ready claims.

Keep this log open until source reviewers replace placeholders with verified notes. Full triage lives in the local source-gap document, with resolution steps in the source-gap resolution workflow.

Status: open · Severity: high

Profile source dates

All 10 project profiles need official source references and last-reviewed dates before claims expand.

Status: open · Severity: high

Article examples

Each brief needs source notes before examples can become publishable educational walkthroughs.

Status: open · Severity: critical

Risk wording

Wallet, payment, DeFi, and retention examples need custody, approval, failure, recovery, and privacy context.

Status: open · Severity: medium

Metadata parity

Article index and detail pages must stay aligned with `article-review-metadata` slugs and draft status.

Step 1 · Intake

Start from an issue ID

Map the draft, profile, or article scaffold to a visible `SG-*` issue before changing copy.

Step 2 · Evidence

Record source labels only

Use local labels such as official docs reviewed, technical reference reviewed, or security note missing.

Step 3 · Decision

Keep holds until complete

Update the decision log only after source notes, risk context, and accessibility evidence are ready.

Batch A · Shared article blockers

Plan parity first

Resolve article source notes, metadata parity, decision-log parity, and manual evidence together before editing one draft.

Batch B-D · Topic blockers

Group by risk context

Review wallet/session controls, payment/audit records, and DeFi supervision as separate batches with local evidence labels.

Batch E-F · Profile and QA

Keep placeholders visible

Profile source dates and manual accessibility evidence remain open until recorded in the local batch plan.

Article Decision Matrix

Hold, defer, or review drafts based on local evidence.

The full local matrix lives in ARTICLE_DECISION_MATRIX.md and the blocker checklist lives in DRAFT_PROMOTION_BLOCKERS.md.

Default state

Hold

Use when source placeholders, risk notes, accessibility evidence, or source-gap lifecycle checks are incomplete.

Explicit exception

Defer

Use only when a reviewer records the postponed source gap and keeps the article in draft status.

Evidence complete

Reviewed

Use only after source notes, risk framing, manual accessibility evidence, copy safety, and local checks pass.

Article Hold Reason Matrix

Why each draft remains blocked from publication.

The full local matrix lives in ARTICLE_HOLD_REASON_MATRIX.md. Hold reasons are editorial notes, not project ratings.

PD-001 · SG-002 SG-003 SG-007 SG-008 SG-009 SG-010

Agent wallet budget

Custody, policy, simulation, emergency pause, metadata parity, manual evidence, and evidence consistency remain incomplete.

PD-002 · SG-002 SG-004 SG-007 SG-008 SG-009 SG-010

Stablecoin API payment

Refund, fee, payment proof, failed settlement, metadata parity, manual evidence, and evidence consistency remain incomplete.

PD-003 · SG-002 SG-005 SG-007 SG-008 SG-009 SG-010

Guarded DeFi rebalance

Slippage, approval scope, oracle assumptions, asset-neutral wording, metadata parity, manual evidence, and evidence consistency remain incomplete.

PD-004 · SG-002 SG-003 SG-007 SG-008 SG-009 SG-010

Wallet session revocation

Permission scope, revocation path, stale-session handling, recovery evidence, metadata parity, manual evidence, and evidence consistency remain incomplete.

PD-005 · SG-002 SG-006 SG-007 SG-008 SG-009 SG-010

Payment audit trail

Retention policy, reviewer roles, sensitive-data limits, incomplete-log handling, metadata parity, manual evidence, and evidence consistency remain incomplete.

Static Review Templates

Reusable checks for sources, articles, and accessibility.

Use these local templates before moving placeholder content into publication-ready copy.

Profile source review

  • Official docs or technical reference reviewed.
  • Repository, standard, audit, or security note recorded.
  • Last-reviewed date added in plain text.
  • Category and technical role kept neutral.

Article review note

  • Working title and reader takeaway are educational.
  • Claims are traceable to primary or technical sources.
  • Risks appear before examples involving funds or wallets.
  • No token, yield, ranking, or endorsement language remains.
  • Detail-page scaffold links back to the local article index.

Accessibility QA note

  • Keyboard path reaches every interactive control.
  • Focus state is visible against the dark background.
  • Status updates use polite live regions.
  • Reduced-motion preferences do not hide critical state.

Reviewer evidence packet

  • Article slug, decision log ID, and blocking source-gap IDs recorded.
  • Source labels, risk context, accessibility evidence, and validation result captured.
  • Remaining hold reason kept visible until every packet field is complete.
  • Template lives in REVIEWER_EVIDENCE_PACKET.md.

Manual A11y Evidence

Evidence fields for checks automation cannot prove.

Record manual results in MANUAL_A11Y_EVIDENCE_LOG.md before clearing accessibility review blockers.

Keyboard and focus

Record the page, viewport, keyboard path, focus visibility, and any trap or skipped control.

Status and state

Record filter announcements, newsletter feedback, menu expanded state, and reduced-motion behavior.

Article scaffold evidence

Record skip-link, heading-order, local-link, and accessibility-note observations for each draft page.

Manual QA Summary

Current local QA evidence remains pending.

Record the completed pass in MANUAL_QA_SUMMARY.md. This summary does not replace the detailed accessibility evidence log.

Pending

Keyboard and focus

Tab order and visible focus still need local reviewer evidence before accessibility blockers can clear.

Pending

Status and reduced motion

Filter announcements, menu state, newsletter feedback, and reduced-motion behavior still need manual notes.

Pending

Article scaffold pass

All five detail pages still need skip-link, heading-order, local-link, and accessibility-note observations.

Pending

Copy safety and release

Neutral educational wording and the final static release check must be recorded after the last local edit.

Static QA Matrix

Local checks for content, accessibility, and source readiness.

This matrix complements the automated scripts and the manual browser smoke test.

Content safety

Search for promotional investment phrasing, unsupported certainty, and examples that could read as advice.

Source completeness

Confirm placeholders remain until official docs, technical references, security notes, and dates are ready.

Hold-state validation

Confirm draft, hold, pending checklist, source placeholders, and decision log references agree everywhere.

Interaction access

Check menu, filters, links, and form feedback with keyboard, focus visibility, and status announcements.

Static asset boundary

Keep the MVP self-contained with local HTML, CSS, and JavaScript only unless a future review approves more.

Static Release Checklist

Release only after local evidence and checks agree.

The full release gate lives in STATIC_RELEASE_CHECKLIST.md.

Run local checks

Record `npm run check` after final edits and keep failures as publication blockers.

Verify static boundary

Confirm no external asset, analytics, live form action, wallet endpoint, payment endpoint, or affiliate URL was added.

Confirm state parity

Compare homepage metadata, article index cards, detail metadata, hold matrix, and publication decision log.

Preserve no-advice framing

Keep drafts educational and remove advice, token, yield, ranking, endorsement, guarantee, or production-runbook language.

Browser Smoke Test

Manual checks for layout, keyboard access, and no-advice copy.

Use this checklist in any modern browser. It is designed for reviewers who do not have browser automation.

  1. Responsive layout: open the page around 1280px, 920px, and 375px wide; confirm content stays readable and the mobile menu opens.
  2. Keyboard path: tab from the address bar through navigation, filter buttons, article links, newsletter input, and footer link.
  3. Visible state: confirm focus outlines, filter pressed states, newsletter status text, and reduced-motion behavior are visible.
  4. Educational framing: scan project profiles and article briefs for source context, risk language, and absence of token or yield advice.
  5. Review templates: compare profile and article edits against the static templates before replacing placeholder language.
  6. Static validation: run the local checks and confirm no external font, image, script, or stylesheet dependency was introduced.
  7. Evidence log: record reviewer initials, viewport, keyboard path, focus evidence, status messages, and follow-up actions locally.

Glossary

Terms readers should understand before comparing projects.

Definitions are simplified for education and should be verified against primary sources.

Agent wallet

A blockchain account controlled through software policies, delegated permissions, and human oversight.

Session key

A scoped signing permission that can be limited by time, budget, contract, or action type.

Stablecoin rail

A payment path that uses a stable asset for settlement while still requiring network and issuer risk review.

Attestation

A signed claim about identity, authorization, or history that another system can inspect.

Simulation

A pre-execution preview used to detect balance changes, approvals, fees, and likely failures.

Revocation

The ability to withdraw permissions, pause activity, or rotate credentials after risk changes.

Builder Playbook

Before giving an agent money, give it constraints.

Budget

Set daily, per-action, and per-counterparty limits. Treat every missing limit as a production bug.

Permissions

Use scoped keys, allowlists, session permissions, and approvals for anything that moves funds.

Observability

Store prompts, tool calls, payment intents, transaction hashes, and policy decisions for audits.

Recovery

Design pause, revoke, refund, and human escalation paths before connecting the first wallet.

Publication Readiness

Checks to complete before treating this MVP as public research.

Keep this list visible so future edits do not drift into advice, unsupported claims, or inaccessible UI.

Weekly Brief

Read educational notes on agent wallets and programmable payments.

Get concise research notes on agent wallets, stablecoin payments, crypto AI tooling, and DeFi automation.