AgentRail — Programmable Wallets & Payment Rails for AI Agents
AGENTRAIL mark AGENTRAIL
{{ n.label }} REQUEST ACCESS

Every agent
needs an
economy.

AGENTRAIL agent core
EVERY AI AGENT GETS A WALLET · EVERY ACTION LEAVES A RECEIPT
THE TRAIL · END TO END
{{ s.n }}
{{ s.title }}
{{ s.body }}
EXPLORE
{{ c.n }} {{ c.title }} {{ c.body }} →

How the
rail works.

One path from a funded wallet to a signed receipt. Identity, policy and proof are three stops on the same line — not three products.

01 · WALLET
HUMAN_019 $100 USDC · FUND AGT-042 ACTIVE · RESEARCH_AGENT POLICY P-04 ≤ $0.05 · AUTONOMOUS WHITELIST · DAILY ≤ $20 ✓ MODEL_07 · $0.034 GPU_12 · $2.410 DATA_04 · $0.900

A funded account,
not an API key.

The owner funds the agent once. From that moment the agent transacts under its own address, and the owner's card is never exposed to a model, a provider or a sub-agent.

ACCOUNTERC-4337 SMART ACCOUNT
ASSETUSDC
KEYSSESSION-SCOPED
02 · POLICY
REQUEST
{{ sel.amt }}
{{ sel.target }}
GATE
ROUTE
{{ sel.verdict }}
{{ sel.rule }}
POLICY {{ sel.policy }}LATENCY {{ sel.latency }}STATUS {{ sel.status }}
{{ r.k }} {{ r.v }} {{ r.mode }}
03 · RECEIPT

Proof, not logs.

Both sides sign. The policy that allowed the spend is attached. The whole thing is anchored in the next block, so an auditor can replay a decision nobody was watching.

SIGNED BY AGENT✓
COUNTERSIGNED BY PROVIDER✓
POLICY PROOF ATTACHED✓
ANCHORED ON-CHAIN✓
AGENTRAIL #92812
FROMAGT-042
TOGPU-091
POLICYP-091
TIME14:32:08.019
AMOUNT2.410 USDC
SETTLED✓
RECEIPT
AGENT
PROVIDER
AMOUNT
POLICY
STATE
{{ l.id }}
{{ l.agent }}
{{ l.provider }}
{{ l.amount }}
{{ l.policy }}
{{ l.state }}

The machine economy
is already moving.

Every rail below is a live payment channel between autonomous parties. No invoices, no net-30, no human in the loop unless policy demands one.

LIVE
{{ s.k }}
{{ s.v }}
AGENT_A ORCHESTRATOR BAL 58.80 USDC GPU_12 $2.410 DATA_07 $0.900 MODEL_01 $0.034 AGENT_B · SUB DELEGATED 12.00 USDC API_04 $0.012 AGENT_C SPAWNED 00:04
{{ f.t }}{{ f.from }}→{{ f.to }}{{ f.amt }}{{ f.hash }}✓
PROVIDERS
{{ p.name }}
STAKE{{ p.stake }}
UPTIME{{ p.uptime }}
REPUTATION{{ p.rep }}
STATUS{{ p.status }}
DISPUTE SLASH RECEIPT MISMATCH FILED STAKE BURNED · RAIL DISCONNECTED

An identity
that can spend.

An agent is not an API key on someone else's card. It is a smart account with an owner, a mandate, a budget and a history it cannot rewrite.

IDENTITY
AGT / 042v1.4
AGENT
RESEARCH_AGENT · ONLINE
WALLET
0x91…A81
OWNER
HUMAN_019
MONTHLY LIMIT
100.00 USDC
DAILY LIMIT
20.00 USDC
AUTONOMOUS
$0.05 / ACTION
SPENT · 30D
41.20 USDC
ROSTER
AGENT
MANDATE
BUDGET
BURN · 24H
STATE
{{ a.id }}
{{ a.mandate }}
{{ a.budget }}
{{ a.burn }}
{{ a.state }}

Docs.

Install the SDK, mint a wallet, attach a policy. The network handles settlement, receipts and provider discovery.

QUICKSTART

01 · INSTALL
npm i @agentrail/sdk
02 · AUTH
export AGENTRAIL_KEY=…
03 · SPAWN
agentrail create-agent
TERMINAL
$ agentrail create-agent
> wallet…………… created
> policy…………… attached
> budget…………… 100 USDC
> network………… online
AGENT_ID
agt_7F291
READY ✓ ▋
SDK · TYPESCRIPT
const agent = await agentrail.create({
budget: "100 USDC",
dailyLimit: "20 USDC",
autoApprove: "< 0.05",
providers: whitelist
});
await agent.pay("model_12", "0.034");
// → receipt #92812 · settled

API REFERENCE

{{ e.method }} {{ e.path }} {{ e.desc }}

CLI

{{ c.cmd }} {{ c.desc }}

WEBHOOKS

{{ h.evt }} {{ h.desc }}

Four layers,
one rail.

Settlement is boring on purpose: stablecoin transfers on an L2, policy enforced at the account layer, receipts anchored in batches.

ARCHITECTURE
{{ l.n }} {{ l.name }} {{ l.body }}
TOKEN
{{ tk.label }}
{{ tk.value }}
{{ tk.body }}
SPEC
{{ s.k }}
{{ s.v }}

Join the
whitelist.

The network is in closed rollout. Agents are spawned in batches once the operator behind them is verified. Tell us what you intend to put on the rail.

I AM JOINING AS
{{ submitNote.text }}
REQUEST QUEUED ✓
You are in the queue. No agent is spawned until an operator signs off.
QUEUE POSITION{{ queuePos }}
OPERATOR{{ form.name }}
CONTACT{{ form.email }}
ROLE{{ form.role }}
REVIEW WINDOW3–5 DAYS
01 · VERIFY
We check the operator behind the agent — entity, jurisdiction, funding source.
02 · SCOPE
A starting policy is drafted with you: caps, whitelist, escalation signer.
03 · SPAWN
Your first agent wallet is minted on testnet with 50 USDC of sponsored budget.
NOTE
Providers go through the same door — stake is only requested after verification.
CLOSED ROLLOUT · BATCH 04

Give your agent
an economic identity.

[ REQUEST AN AGENT → ] READ THE DOCS
VERIFY→SCOPE POLICY→SPAWN ON TESTNET→50 USDC SPONSORED