# Budgets, receipts and independent execution approval

A concrete AI trading integration flow that separates API consumption, unknown receipt recovery and independently approved Testnet orders.

Documentation preview · Execution: Testnet only

Content reviewed: 2026-10-24

Canonical: https://catalystdesk.ai/en/research/agent-budgets-and-approvals

Revision: discovery-2026-10-24.2

## At a glance

A funded API balance authorizes service spending within its limits. It does not authorize a trade. Preserve a durable request ID for paid research and use independent approval for any configured Testnet execution workflow.

## Bound the research request

For a synthetic budget of USD 0.01, set X-CatalystDesk-Max-Cost-Nano to 10000000. Accept a quote only if its amount fits that limit, its input matches your intended coin/dex and it has not expired. This number is a budgeting example, not a published service price.

## Recover uncertainty without a second charge attempt

If the network fails after dispatch, query the saved request ID. A receipt may still be pending or reconciling. Reusing a key does not make it safe to create a new request ID and blindly repeat the action. Do not infer execution or payment from a local success message.

## Approve the exact execution proposal

Where enabled, a Testnet proposal binds context, account and instrument before independent operator approval. The client explicitly invokes execution and later checks receipts. Model output, project ownership and Credits cannot self-approve a proposal. This is a control-flow walkthrough, not evidence of a real venue fill.

## Inspect a structured sample

Illustrative workflow, not market data or a complete API response

Illustrative workflow only. No source data, private model parameters or execution logic are included.

```json
{
  "example": "illustrative workflow only",
  "question": "What could change this market assessment?",
  "evidence": "Source context and missing information are identified.",
  "decision_state": "Human review required",
  "limitations": "No current market observation or execution instruction."
}
```

## Check the service before integration

Check enabled capabilities in the authorized workspace before use. Public documentation does not activate a service. Simulation and testnet prototypes do not establish live trading availability.

## Protocol references

Venue documentation describes source protocols. It does not establish CatalystDesk deployment, integration endorsement or commercial availability.

- [Hyperliquid API: Info endpoint](https://hyperliquid.gitbook.io/hyperliquid-docs/for-developers/api/info-endpoint)
- [Hyperliquid API: WebSocket subscriptions](https://hyperliquid.gitbook.io/hyperliquid-docs/for-developers/api/websocket/subscriptions)

## Continue with the public documentation

- [Market intelligence for AI trading agents](https://catalystdesk.ai/en/solutions/ai-trading-agents)
- [Hyperliquid research and testnet workflows](https://catalystdesk.ai/en/integrations/hyperliquid)
- [Plan your first research workflow](https://catalystdesk.ai/en/developers/quickstart)
- [How to evaluate market context](https://catalystdesk.ai/en/methodology/market-context)
- [Product Credits and usage](https://catalystdesk.ai/en/pricing/api-credits)
- [From market observations to an agent review](https://catalystdesk.ai/en/research/hyperliquid-context-walkthrough)
- [Separate price differences from timing evidence](https://catalystdesk.ai/en/research/basis-and-source-clocks)
- [Understand Hyperliquid BTC and ETH with an AI agent](https://catalystdesk.ai/en/agents/hyperliquid-analysis)
- [Monitor crypto market changes with a bounded agent](https://catalystdesk.ai/en/agents/market-monitoring)
- [Simulate a BTC or ETH trade before taking risk](https://catalystdesk.ai/en/agents/trade-simulation)