PRIVATE BETA · NO CUSTODY OF FUNDS

Your agents can spend money.
Remit decides how much.

One looping agent burned $6,531 on AWS overnight. Another ran up $50K. Remit puts hard caps, human approvals, and a kill switch between your agents and your money — and hands your auditor the receipts.

WORKS WITH CLAUDE CODE · CURSOR · CODEX · LANGGRAPH VIA MCP OR REST
CARDS · ACH · PAID APIS — POLICY ENFORCED BEFORE MONEY MOVES

2 locksfrozen at auth + spend layer
<1 clickto freeze any agent
Σ = 0double-entry by construction
100%movements in append-only log
Built for the audit, not just the bill

"Show me who approved that."
Say it with a link.

SOC 2 evidence packs, AIUC-1, and payment-network mandates all converge on one requirement for autonomous agents: produced, explicable records of every consequential action — especially money. Remit's append-only ledger turns agent spending from an audit liability into an audit asset.

  • ✓Every spend, rule hit, and approval attributed to a named human or an agent key — never just "the company"
  • ✓Immutable double-entry ledger underneath — balances reconcile by construction, not by spreadsheet
  • ✓Tamper-evident event chain — hash-chained, exportable as auditor-readable CSV for any date range
SPEND EVENTS BY HOUR — scrape-agent
Tuesday. Notice anything?
00:00a normal workday…23:00
19:00–23:00: retry loop, ~340 identical calls/hour. Remit's daily cap tripped at 19:04. The freeze happened before finance noticed. Nobody paged anyone.
How it works

Set rules once. Every agent obeys.

Policy lives on the server, not in the prompt. Agents can't negotiate with their own limits.

01 — IDENTITY

Add an agent

Mint an ak_ key bound to one agent, enforced server-side. Agents get granted access to wallets — they never own them.

02 — POLICY

Set the rules

Per-transaction, daily, monthly caps. Allowed and blocked categories. Above $X, a human approves before money moves.

03 — OVERSIGHT

Watch — or kill

Every decision in one feed. Something wrong? Freeze kills the agent at the auth layer and again at the spend layer. Two locks, one click.

Rail-agnostic on purpose

The neutral layer above every rail.

Cards, bank transfers, paid APIs, x402, stablecoins — Remit governs them all without touching custody of your funds. That neutrality is the point: no rail owns us, so our incentives stay aligned with yours.

CARDSACH / RTPPAID APIs X402STABLECOINSMCP-NATIVE
Questions engineers ask us
Do I need to change how my agent pays?

No. Route payment calls through the Remit API or MCP tools. Policy is evaluated before anything moves, and required idempotency keys mean a retry storm can't double-spend.

What stops a compromised agent key?

Agent keys carry no wallet ownership — only grants. Freeze rejects at authentication (403) and again at spend time. Every action requires an X-Instance-ID, so even a stolen key leaves attributable forensics.

Are you a payment provider?

Deliberately not. Remit never holds or transmits funds — it governs them. That keeps us neutral across rails and keeps you free of new counterparty risk.

What does my auditor actually receive?

A date-range activity export: every movement, its policy outcome, and the named human or agent key behind each decision — hash-chained so tampering is detectable. Attach it to your SOC 2 evidence pack.

Private beta

Give your agents a budget.
Keep your audit clean.

LIMITED SEATS · YOUR FEEDBACK SHAPES THE ROADMAP · HELLO@REMITAPP.DEV