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

# Track Pump.fun Token Mints with Parsed Streams

> Subscribe to Pump.fun's create and create_v2 instructions with Parsed Streams and build a reconnect-safe listener that logs every new token mint, creator, name, symbol, and URI.

This guide builds a long-running listener for **every new Pump.fun token**. It filters on the two instructions Pump.fun uses to launch a token, and stays connected across idle timeouts and deploys.

<Steps>
  <Step title="Look Up the Program">
    Pump.fun launches tokens through two instructions depending on version: `create` and `create_v2`. Confirm the exact names and account roles with [`describeProgram`](/docs/parsed-streams/quickstart#discovery) before filtering on them:

    ```json Request theme={"system"}
    { "jsonrpc": "2.0", "id": 1, "method": "describeProgram", "params": [{ "program": "6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P" }] }
    ```

    Check the response's `instructions` list for `create` and `create_v2`, and its `roles` list for the account you want — typically the new **mint** address, and either a `creator` account role or a `creator` field in `args`. The two instruction versions don't necessarily share a shape, which is why the code below checks both an arg and a couple of role names rather than assuming one.
  </Step>

  <Step title="Build the Filter">
    Filter on the program and both instruction names. Leave `includeCpi` on — Pump.fun's `create`/`create_v2` are usually top-level, but this is cheap insurance — and exclude failed transactions since a failed deploy never produces a live mint:

    ```json theme={"system"}
    {
      "programs": ["6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P"],
      "instructionNames": ["create", "create_v2"],
      "includeFailed": false,
      "includeCpi": true
    }
    ```
  </Step>

  <Step title="Connect with a Reconnect-Safe Client">
    A deploy tracker is a long-running process, so treat disconnects as routine rather than exceptional: watchdog the initial handshake, keep the connection alive on a quiet filter, and reconnect automatically on close.

    ```javascript pumpfun-deploys.js theme={"system"}
    const API_KEY = process.env.HELIUS_API_KEY;

    if (!API_KEY) {
      console.error("Missing HELIUS_API_KEY.");
      process.exit(1);
    }

    const URL = `wss://<ENDPOINT>/?api-key=${API_KEY}`;
    const PUMP_PROGRAM = "6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P";

    const SUBSCRIBE_ID = 1;
    const PING_ID = 99;
    const SUBSCRIBE_REQUEST = {
      jsonrpc: "2.0",
      id: SUBSCRIBE_ID,
      method: "parsedTransactionSubscribe",
      params: [
        {
          programs: [PUMP_PROGRAM],
          instructionNames: ["create", "create_v2"],
          includeFailed: false,
          includeCpi: true,
        },
      ],
    };

    const CONNECT_TIMEOUT_MS = 10_000;
    const PING_INTERVAL_MS = 30_000; // server closes idle connections after 10 min

    const ts = () => new Date().toISOString();

    function connect() {
      console.log(`[${ts()}] connecting…`);
      const ws = new WebSocket(URL);
      let pingTimer = null;

      // Watchdog: if the handshake never completes, force-close so we retry
      // instead of hanging silently in CONNECTING.
      const connectTimer = setTimeout(() => {
        if (ws.readyState === WebSocket.CONNECTING) {
          console.error(`[${ts()}] handshake stalled after ${CONNECT_TIMEOUT_MS}ms — closing`);
          ws.close();
        }
      }, CONNECT_TIMEOUT_MS);

      ws.addEventListener("open", () => {
        clearTimeout(connectTimer);
        console.log(`[${ts()}] connected, subscribing to pump create/create_v2`);
        ws.send(JSON.stringify(SUBSCRIBE_REQUEST));
        pingTimer = setInterval(() => {
          if (ws.readyState === WebSocket.OPEN) {
            ws.send(JSON.stringify({ jsonrpc: "2.0", id: PING_ID, method: "ping" }));
          }
        }, PING_INTERVAL_MS);
      });

      ws.addEventListener("message", (event) => handleMessage(event.data));

      ws.addEventListener("error", (err) => {
        console.error(`[${ts()}] socket error:`, err.message || err);
      });

      ws.addEventListener("close", (event) => {
        clearTimeout(connectTimer);
        if (pingTimer) clearInterval(pingTimer);
        console.log(`[${ts()}] closed (code=${event.code}, reason="${event.reason}"). reconnecting in 3s…`);
        setTimeout(connect, 3000);
      });
    }

    connect();
    ```

    <Tip>
      This uses a fixed 3-second reconnect delay for simplicity. For a production listener, back off exponentially instead — see [Handling Reconnects](/docs/parsed-streams/guides/handling-reconnects).
    </Tip>
  </Step>

  <Step title="Handle Notifications and Log Each Deploy">
    Route incoming messages by `id` (your own requests) or `method` (server-pushed notifications), then pull the mint, creator, and metadata out of the decoded instruction:

    ```javascript theme={"system"}
    // Pull an account pubkey out of a decoded instruction by its role name.
    function accountByRole(decoded, role) {
      return decoded?.accounts?.find((a) => a.name === role)?.pubkey ?? null;
    }

    let deployCount = 0;

    function handleDeploy(tx, ix) {
      deployCount += 1;
      const args = ix.decoded?.args ?? {};
      const mint = accountByRole(ix.decoded, "mint");
      const creator =
        args.creator ??
        accountByRole(ix.decoded, "creator") ??
        accountByRole(ix.decoded, "user");

      console.log(`[${ts()}] pump deploy #${deployCount} (${ix.instructionName})`);
      console.log(`    mint:    ${mint}`);
      console.log(`    creator: ${creator}`);
      console.log(`    name:    ${args.name ?? "?"}`);
      console.log(`    symbol:  ${args.symbol ?? "?"}`);
      console.log(`    uri:     ${args.uri ?? "?"}`);
      console.log(`    slot=${tx.slot} sig=${tx.signature}`);
    }

    function handleMessage(raw) {
      let msg;
      try {
        msg = JSON.parse(raw);
      } catch {
        console.log(`[${ts()}] non-JSON message:`, raw);
        return;
      }

      if (msg.error) {
        console.error(`[${ts()}] error (code=${msg.error.code}): ${msg.error.message}`);
        return;
      }
      if (msg.id === SUBSCRIBE_ID && typeof msg.result === "number") {
        console.log(`[${ts()}] subscribed (subscription id=${msg.result})`);
        return;
      }
      if (msg.id === PING_ID) return; // keepalive pong, ignore quietly

      if (msg.method === "parsedTransactionNotification") {
        const { value } = msg.params.result;
        // Default details is "full", so value.instructions is the whole
        // transaction; matchedIndexes points at just the create/create_v2
        // instructions this filter hit.
        for (const i of value.matchedIndexes) handleDeploy(value.transaction, value.instructions[i]);
        return;
      }

      console.log(`[${ts()}] message:`, JSON.stringify(msg, null, 2));
    }
    ```

    Because `includeFailed` is `false`, every notification you log is a token that actually deployed. Check `ix.decoded` is present before trusting `args` and `accounts` — an unrecognized build of the Pump.fun program arrives with `decoded: null` and would otherwise show up as `mint: null`.
  </Step>
</Steps>

## Next Steps

<CardGroup cols={2}>
  <Card title="Track Jupiter Swaps" icon="arrow-right-arrow-left" href="/docs/parsed-streams/guides/track-jupiter-swaps">
    The describeProgram → build filter → subscribe flow in more detail.
  </Card>

  <Card title="Handling Reconnects" icon="rotate" href="/docs/parsed-streams/guides/handling-reconnects">
    Exponential backoff and slot-gap backfill for long-running listeners.
  </Card>

  <Card title="Fetch Pump.fun Mints" icon="clock-rotate-left" href="/docs/parsed-events/guides/fetch-pumpfun-mints">
    The historical version: page through a creator's past deploys with Parsed Events.
  </Card>
</CardGroup>