> ## Documentation Index
> Fetch the complete documentation index at: https://ti-mm-mycompc.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Solana priority fees that stop agents overpaying

> Sample recent priority fees and return low, medium, and high tier recommendations with plain-English advice so agents pay the right amount every time.

Solana priority fees change every 10 seconds. Agents that hardcode a fee either fail during congestion or overpay by 3x during quiet periods. This endpoint samples the current network, buckets fees into low/medium/high tiers, and returns plain-English `llm_advice` your model can act on directly.

<Note>
  Paid endpoint. Pass your key in the `x-api-key` header. See [Get an API key](/guides/claim-api-key).
</Note>

## Why agents use this

* **Fees move constantly.** A static value is either wasteful or broken. There's no correct hardcoded number.
* **LLM-native output.** `llm_advice` is a plain sentence like "Network is quiet, use low tier to save 60%." Feed it straight into your prompt.
* **Three-tier structure matches user intent.** "Get me in fast" (high), "reasonable" (medium), "I'm patient" (low). Match to user preference or agent urgency.
* **Live congestion signal.** `current_congestion` is a stable classifier your bot can log and monitor over time.

## Use cases

* **Dynamic fee per trade.** Loop your bot: call `optimal-fee`, use the tier that matches urgency, submit. Typical savings: 30-50% versus a fixed conservative fee.
* **Time mints to congestion dips.** Poll every minute; auto-mint the moment `current_congestion` flips to `LOW`.
* **CI/monitoring alert.** Slack notification when fees spike above a threshold so ops knows to pause automated batches.
* **Per-user fee preference in a DEX UI.** Show the three tiers with `description` and `estimated_time` from the response. Users pick, bot executes.
* **Fee attribution in analytics.** Log the recommended tier at trade time so you can compare theoretical optimum versus what you actually paid.

## Recipe: adaptive fee selection for a trading bot

```typescript adaptive-fee.ts theme={null}
const API_KEY = process.env.API_KEY!;
const GATEWAY = "https://solana-pulse-gateway-1021990235790.us-central1.run.app";

type Urgency = "low" | "medium" | "high";

async function selectPriorityFee(urgency: Urgency = "medium") {
  const res = await fetch(`${GATEWAY}/api/solana/optimal-fee`, {
    headers: { "x-api-key": API_KEY },
  });
  const { current_congestion, tiers, llm_advice } = await res.json();

  const tier = tiers[urgency];
  console.log(`Congestion: ${current_congestion}. ${llm_advice}`);
  console.log(`Using ${urgency}: ${tier.lamports} lamports (${tier.estimated_time})`);

  return tier.lamports;
}

// In your tx builder:
const priorityFee = await selectPriorityFee("high"); // sniper mode
```

<Tip>
  Combine with [Blockhash](/api-reference/solana/blockhash) to build fully-configured transactions client-side without your own RPC. Two calls, no infra.
</Tip>