> ## Documentation Index
> Fetch the complete documentation index at: https://docs.prflght.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Quick Start: Send Your First Protected Transaction

> Connect your wallet, set your agent policy, install the SDK, and send your first firewall-protected Solana transaction in under ten minutes.

Prflght protects your AI agent's transactions through a four-stage pipeline: simulation, policy evaluation, protocol health check, and on-chain attestation. This guide walks you through every step — from creating your account to sending your first protected transaction.

<Steps>
  <Step title="Connect your wallet">
    Go to [app.prflght.xyz](https://app.prflght.xyz) and connect your Solana wallet. Prflght supports **Phantom** and **Solflare**. Once connected, sign the authentication message in your wallet. No password is required — your signature proves ownership of the wallet.
  </Step>

  <Step title="Copy your API key">
    After authenticating, open the **API Keys** section of the dashboard and copy your API key. Keep this value secure — it authorizes all firewall checks made on behalf of your agent.
  </Step>

  <Step title="Set your agent policy">
    Open the **Policy editor** on the dashboard and configure the rules for your agent:

    * **Max notional** — the maximum USD value of a single transaction
    * **Allowed programs** — a list of Solana program IDs your agent may interact with
    * **Slippage limit** — the maximum acceptable slippage percentage
    * **Daily transfer limit** — the cumulative USD value allowed per day

    Save the policy. Prflght applies it to every transaction your agent submits.
  </Step>

  <Step title="Install the SDK">
    Add `@prflght/sdk` to your project:

    <CodeGroup>
      ```bash npm theme={null}
      npm install @prflght/sdk
      ```

      ```bash bun theme={null}
      bun add @prflght/sdk
      ```
    </CodeGroup>
  </Step>

  <Step title="Initialize the Firewall client">
    Import `Firewall` and construct a client with your API URL and agent ID. The `agentId` is your agent's wallet public key as a base58 string.

    ```typescript theme={null}
    import { Firewall, FirewallDenyError } from '@prflght/sdk'

    const firewall = new Firewall({
      apiUrl: process.env.PRFLGHT_API_URL,   // https://api.prflght.xyz
      agentId: agentKeypair.publicKey.toString(),
    })
    ```
  </Step>

  <Step title="Check your transaction">
    Before you send a transaction, pass its base64-encoded bytes to `firewall.check()`. Prflght simulates the transaction, evaluates it against your policy, and checks protocol health.

    ```typescript theme={null}
    const result = await firewall.check(txBase64)
    ```

    If the transaction passes all checks, `result.instructions` contains the attestation instructions you must prepend to your transaction.

    <Note>
      You must prepend `result.instructions` to your transaction's instruction list. The on-chain program rejects any transaction that does not include a valid attestation.
    </Note>
  </Step>

  <Step title="Prepend attestation instructions and send">
    Build the final transaction with the attestation instructions first, then your own instructions, and submit it.

    ```typescript theme={null}
    import { Firewall, FirewallDenyError } from '@prflght/sdk'

    const firewall = new Firewall({
      apiUrl: process.env.PRFLGHT_API_URL,
      agentId: agentKeypair.publicKey.toString(),
    })

    try {
      const result = await firewall.check(txBase64)
      // Prepend attestation instructions before your own
      const finalTx = buildTxWithIxs([...result.instructions, ...yourIxs])
      await sendAndConfirm(finalTx)
    } catch (err) {
      if (err instanceof FirewallDenyError) {
        console.error('Blocked:', err.reason) // e.g. "notional exceeds policy limit"
      }
    }
    ```

    `FirewallDenyError` is thrown when Prflght blocks the transaction. The `err.reason` field explains which check failed — use it to adjust your policy or debug your agent's behavior.
  </Step>
</Steps>