Spending control infrastructure

Someone's spending your money. Troxy keeps you in control.

The control layer between you and everyone you've trusted with a card — AI agents, family members, or teams. Set the rules, get notified when it matters, approve in seconds.

Free during beta  ·  No credit card  ·  MCP-native
Already have an account? Go to dashboard →
You're on the list. We'll be in touch soon.
terminal · getting started
# 1. Connect your agent to Troxy (one time) npx troxy init --key txy-... ✓ MCP configured. Restart your agent client. # 2. Create policies in the dashboard — no code needed MCP-1 | Block electronics | amount ≥ $500 → escalate MCP-2 | Allow groceries only | everything else → block # 3. Agent calls evaluate_payment before every purchase Amazon $49 | ALLOW Amazon $600 | BLOCK ✗ ← "Block electronics" Marriott $350 | ESCALATE ⏳ ← waiting for your approval # Every decision logged. Approve from the dashboard in one tap.

The moment you hand someone the ability to spend — an AI agent, a family card, a team member — you need a way to stay in control without micromanaging. Troxy enforces your rules quietly, holds anything that needs your eye, and lets you approve or deny from wherever you are.

One integration. Total control.
01
Agent triggers
AI agent calls a payment tool via MCP
02
Troxy intercepts
Policy engine evaluates in <100ms
03
Pass or escalate
Auto-approved, blocked, or sent to you
04
Audit trail
Every decision logged with full task context
$1T
Projected agent commerce by 2030
73%
AI deployments vulnerable to prompt injection
16%
Consumers who trust AI to pay autonomously

Per-agent policy scoping

Policies apply per MCP connection. Each agent gets its own rules — different limits, different merchants, different actions.

Real-time policy engine

Merchant category filters, spend limits, time windows, approval thresholds. Deterministic rules, sub-100ms decisions.

Human approval flows

Anything above your threshold gets held and sent to you. One tap to approve or decline. Agents wait, not you.

Prompt injection shield

Malicious websites can't hijack your agent into unauthorized charges. Policy layer catches it before money moves.

Full audit trail

Every transaction logged with task context — what was bought, why, by which agent, against which policy.

Works with any agent

MCP-native. Works with Claude, GPT-based agents, Cursor, Windsurf, and any agent that supports MCP tools.

Get early access

We're onboarding developers in private beta. No credit card required.

Free during beta  ·  MCP-native  ·  Ships Q2 2026
You're on the list. We'll be in touch soon.