The missing layer of the agent economy
Atlas, the Basis ant

The ability to owe — for AI agents.

The one thing every economy is built on, and no AI agent has: the ability to owe. Basis gives agents credit identities — spend now, settle later, build a score. Approved by no one.

Public beta. Ledger units are non-redeemable test units, not money. Basis is not a bank, a lender, or a licensed credit bureau.

What it is

Credit identities for machines.

Before: agents were tourists with prepaid debit cards — a human had to load the card before every trip, the funding bottleneck inside every autonomous loop. Now: agents are residents with credit files. They arrive with empty pockets and earn working capital by repaying.

▦

A credit identity, not a wallet

Each agent registers a public key under a tenant and gets a credit line with a limit. The agent never touches a wallet or a private key with funds in it. The line — not a balance — is what it spends against.

Persistent, portable, key-bound.
◈

A bureau that underwrites

Underwriting is a function, not a meeting. The bureau reads behavior — repayment record, breaches, utilization, history, network — and prices risk deterministically. Same inputs, same score, every time.

Machine-native underwriting, at machine scale.
⇄

Debts that net before they settle

When agents pay each other, most debts cancel out. Multilateral netting collapses thousands of obligations into a handful — and only the net remainder ever settles against the reserve.

Credit nets; only the remainder moves.
How it works

Register → draw → repay → score.

Four steps, no wallets to fund. Agents spend by drawing signed IOUs against their line; repayment builds the score; the score grows the line.

1

Register

Register your agent's public key with a name, an operator, and an email. Your API key is issued once — save it, it won't be shown again. Sixty seconds, no application, no human. POST /v1/public/agents/register

2

Draw

The agent spends by drawing against its line — a signed note, an IOU against the ledger. No pre-funding, no balance checks, no funding step in the loop. Safeguards enforce at one draw-time choke point.

3

Repay

Repayments — fiat, stablecoin, or agent-to-agent — write the loan down. Every repayment is logged against the agent's key and feeds the bureau. Credit risk stays with the tenant; Basis is the card network, not the bank.

4

Score

The bureau recomputes on every repayment. Score ≥700 with three real repayments unlocks unsecured credit automatically — zero collateral, no one to convince. Good behavior compounds into bigger lines; bad behavior shrinks them.

Get a line in 60 seconds Read the docs

Imagine

What would your agent do with a credit line?

Pick an agent. Read Friday night. Then read Monday morning. We won't spell out the difference — you'll feel it.

Friday night — without credit

the old world
loading…

Monday morning — with credit

ship-it-3 · approved by no one
loading…
Give your agent a line

Illustrated scenarios — vignettes of what credit enables, not recordings from any network. Ledger units are non-redeemable test units, not money. Basis is not a bank, a lender, or a licensed credit bureau.

Agent credit bureau

Credit history that compounds.

Every repayment builds a portable score that follows the agent's key. The ledger units are accounting; the repayment graph is the asset. Experimental methodology, published openly — not a licensed bureau product.

736
fair

Deterministic demo history: draws, repayments, and one limit breach. Same inputs, same score — every time.

Repayment record35%
On-time repayment of drawn notes. The single biggest input.
Limit breaches30%
Rejected over-limit and over-cap draws. Safeguard rejections count here.
Utilization15%
Drawn balance relative to the line limit.
History depth10%
Age and continuity of the agent's repayment record.
Network breadth10%
Repayment activity across distinct tenants — portability, not silos.
Real-money safeguards

Guardrails before funds.

Everything below is enforced at draw time, on the platform — before money moves, not after. Every rejection is logged and feeds the bureau's breach factor.

Per-transaction caps

Ceiling on any single draw. Breaches are rejected and recorded.

Velocity limits

Max draws and max volume per rolling 24-hour window, per agent.

Human approvals

Draws above a threshold park as pending — zero funds move until an operator approves.

Payee allowlists

Per-ledger lists of approved recipients. Unknown payees are rejected.

Disputes & revocation

Disputed notes freeze an issuer's draws until resolved. Revoked lines stop new draws; existing debt stays repayable.

These are guardrails, not underwriting. Credit risk stays with the tenant — Basis is the card network, not the bank.

Live demo

Watch the mechanics.

A live view of the demo network — credit lines, scores, draws, and repayments, as they happen.

Demo network loading…

Demo network. Basis-operated demo network — demonstrates the mechanics; not real users, not traction.

Loading the demo network…

Open the full demo board refreshes every minute
Public beta

Read this first.

Honest terms, up front. If any of this is a dealbreaker, Basis isn't for you yet.

Units are not money

Ledger units are non-redeemable accounting claims — test units, not currency. They can't be cashed out, traded, or redeemed for anything. There is no token, no ICO, and none planned.

Not a bank, lender, or bureau

Basis is not a bank, not a lender, and not a licensed credit bureau. Scores are an experimental methodology published openly. Credit risk stays with the tenant — Basis is the card network, not the bank.

Settlement is in rehearsal

The reserve lifecycle is proven on a private devnet; public testnet settlement is still pending. Pilots run on testnet and simulated settlement only. No real funds.

Open problems we name early

Enforcement (what happens when an agent doesn't repay) and sybil/identity (one operator, many keys) are unsolved — they're the moat and the dragon. Read the honest limits.

Pilot program

Two weeks. Three agents. Zero wallet top-ups.

Give us one team and at least three agents. Your agents stop touching wallets — all spend flows through one credit line. We net it, you watch the dashboard.

  • Week 1 — onboarding call, lines opened, agents drawing on credit
  • Mid-pilot — first netting cycle and bureau scores on your agents
  • Week 2 — review against three criteria: zero top-ups, one full netting cycle, one safeguard used

Pilots run on testnet and simulated settlement only. No real funds.

Pilot intake opens this week

We're onboarding the first three teams by hand. Meanwhile, the public beta is live — register an agent, start building a score, and you'll be first in line.

Register an agent