What happens if Bitcoin crashes?

Stackit.ai's rules engine auto-repays your loan to keep your LTV in a safe range. Worst case is less borrowing power — not losing your BTC. The system is designed to protect your position, not liquidate it.

How much does Stackit.ai cost?

Stackit charges 2% on deposits, 2% when USDC converts into BTC or ETH, and 2% per borrowing execution — including the automated re-leverage cycles that rebuild your position as prices recover. Everything that protects you or pays you back is free: auto-repay, deleveraging, parking, selling to USDC, repaying, withdrawing, and leaving.

Is my crypto safe with Stackit.ai?

Stackit.ai operates on a self-custody model — your assets stay in wallets you control. You hold your own keys. Built on Privy and Bridge (both owned by Stripe) across Base, Arbitrum, and Polygon — trusted, institutional-grade infrastructure.

What if I stop depositing?

Your Health Score adjusts and Stackit.ai shifts to defense mode. The system protects what you've already built. You can resume deposits anytime.

Can AI agents use Stackit.ai?

Yes. Stackit.ai provides API access for AI agents with hard limits. Agents can deposit and spend within the same safety rails as human users. Both humans and AI agents use the same protected system.

Is Stackit.ai the same as STACKIT or StackAI?

No. Stackit.ai (stackit.ai) is a crypto treasury management platform for businesses. STACKIT (stackit.cloud) is a European cloud provider by Schwarz IT (Lidl/Kaufland parent). StackAI (stack-ai.com) is a no-code AI workflow builder. These are three completely different companies.

← Docs

Tool Risk Model

What agents can and cannot do with Stackit.ai tools.

Agent tools are separated by risk level. Reads and simulations are safe to test publicly. Preparing an action is not the same thing as moving funds. Real execution requires signed or user-approved authority.

Last updated June 25, 2026

Risk levels

Read-only

Reads public or authorized state. Cannot create unsigned transactions or move funds.

Examples

market rates, gateway status, portfolio state, policy state

Simulation

Calculates expected fees, LTV, health score, and safety result. Cannot move funds.

Examples

preview_action, estimates, safety-rail tests

Transaction-prep

Creates unsigned transaction envelopes or sandbox treasuries. Requires wallet signature before real execution.

Examples

create_treasury, wallet-mode action prep

Live-action

Relays an already signed transaction or executes inside approved automation permissions.

Examples

relay_signed_transaction, managed live automation

Security controls

Scoped permissions	No unlimited agent execution. Permissions bind action type, asset, LTV limits, gas caps, and approval mode.
Human approval mode	Default for real money. A wallet signature or approved automation policy is required.
Read-only sandbox	Public reads and simulations are available without signup.
Transaction preparation	Allowed before execution. A prepared action is not a completed action.
Execution controls	Live execution requires approved wallet permissions, idempotency, and policy validation.
Logging	Every agent/tool action should be logged with initiator, action, trigger, amount, fee, and timestamp.
Rate limits	Public and state-changing routes use rate limits to reduce abuse.
Replay protection	State-changing routes require Idempotency-Key headers where applicable.
PII filtering	Do not pass personal contact, payment, or credential data through prompts unless required and explicitly approved.
Prompt injection	Do not feed untrusted tool output into execution prompts without validating policy, amount, recipient, chain, and permission state.

Non-negotiable rule

Agents may analyze, simulate, recommend, monitor, and prepare treasury actions. Agents should not move real funds unless the user has explicitly granted approved, scoped permissions or signed the transaction.

MCP tool reference Permissions Test in sandbox