メインコンテンツへスキップ

Overview

LaserStream is a managed Solana gRPC streaming service. It is wire-compatible with the open Yellowstone gRPC protocol — so any Yellowstone client works out of the box — and adds production features like historical replay, multi-node failover, and a fully managed environment. LaserStream uses the open source gRPC protocol, ensuring no vendor lock-in and maximum compatibility with existing gRPC implementations. You can connect either with the standard @triton-one/yellowstone-grpc client or use the performance-optimized Helius LaserStream SDK for added benefits including higher throughput, automatic reconnects, subscription management, error handling, and more.

LaserStream SDK is 40x Faster vs. JavaScript Yellowstone Clients

Learn how we used Rust Core with zero-copy NAPI bindings to maximize JavaScript SDK performance
Performance Notice: If you experience any lag or performance issues with your LaserStream connection, please refer to the Troubleshooting section for common causes and solutions.
No Compression: To minimize latency, LaserStream does not compress gRPC response messages. Setting Accept-Encoding with gzip or zstd will have no effect — responses are always returned uncompressed.

Endpoints & Regions

LaserStream is available in multiple regions worldwide. Choose the endpoint closest to your application for optimal performance:

Mainnet Endpoints

Devnet Endpoint

Network & Region Selection:
  • For production apps, pick the mainnet endpoint nearest your server for best performance (e.g., if deploying in Europe, use Amsterdam (ams) or Frankfurt (fra))
  • For testing, use: https://laserstream-devnet-ewr.helius-rpc.com.

Log Truncation

By default, LaserStream truncates transaction log messages to 10 KB for better speed and performance. If you need full logs, dedicated untruncated endpoints are available — see Log Truncation.

Quickstart

Get started with LaserStream from your Helius Dashboard. Mainnet requires a Business or Professional plan; Devnet is available on Developer and above. See Plans & Pricing for details.
1

Create a New Project

2

Install Dependencies

We use tsx because the default npx tsc --init on TypeScript 5.x sets verbatimModuleSyntax, module: "nodenext", and types: [], which all break a quick ts-node index.ts run. tsx runs .ts files without a tsconfig.
3

Obtain Your API Key

Generate a key from the Helius Dashboard.This key will serve as your authentication token for LaserStream.
Plan Requirements: LaserStream devnet is available on all plans. LaserStream mainnet requires a Business or Professional plan.
4

Create a Subscription Script

Create index.ts with the following:
5

Replace Your API Key and Choose Your Region

In index.ts, update the config object with:
  1. Your actual API key from the Helius Dashboard
  2. The LaserStream endpoint closest to your server location
Network & Region Selection Examples:
  • For Production (Mainnet):
    • Europe: Use fra (Frankfurt), ams (Amsterdam), or lon (London)
    • US East: Use ewr (New York)
    • US West: Use slc (Salt Lake City) or lax (Los Angeles)
    • Asia: Use tyo (Tokyo) or sgp (Singapore)
  • For Development (Devnet):
    • Use https://laserstream-devnet-ewr.helius-rpc.com
6

Run and View Results

Whenever a confirmed token transaction involves TokenkegQfeZyiNwAJbNbGKPFXCWuBvf9Ss623VQ5DA, you’ll see the data in your console.

Common Workflows

Step-by-step guides for the workflows we see most often. Each guide uses the helius-laserstream SDK with auto-reconnect and historical replay built in.

Account Subscriptions

Monitor balance, data, and ownership changes on specific accounts with filters.

Transaction Monitoring

Stream transactions involving target accounts, filter by program, vote, or failure status.

Slot & Block Monitoring

Track network consensus, block production, and commitment-level transitions.

Decoding Transaction Data

Parse the binary transactionUpdate payloads into readable Solana transactions.

Stream Pump AMM Data

Real-world example: monitor Pump AMM trades with reconnect-safe filters.
The @triton-one/yellowstone-grpc client works against the same endpoints if you prefer the raw Yellowstone protocol. See the Yellowstone gRPC reference for protocol-level details.

Subscribe Request

In the subscribe request, you need to include the following general parameters:
Historical Replay: You can optionally include a fromSlot field (a u64 number) in the main SubscribeRequest object to replay data from a specific slot onwards. Replay is currently limited to the last 216,000 slots (≈24 hours); note that replays older than ~20 minutes return finalized-only data.
Next, you’ll need to specify the filters for the data you want to subscribe to, such as accounts, blocks, slots, or transactions.
Define filters for slot updates. The key you use (e.g., mySlotLabel) is a user-defined label for this specific filter configuration, allowing you to potentially define multiple named configurations if needed (though typically one is sufficient).
Define filters for account data updates. The key you use (e.g., tokenAccounts) is a user-defined label for this specific filter configuration.If all fields are empty, all accounts are broadcasted. Otherwise:
  • Fields operate as a logical AND.
  • Values within arrays act as a logical OR (except within filters, which operate as a logical AND).
Tracking more than ~10,000 accounts? Instead of an explicit pubkey list (32 bytes per account), use a compressed cuckoo filter (~3–4 bytes per account) to subscribe to hundreds of thousands of accounts in a single stream. Available in the Rust and JavaScript SDKs.
Define filters for transaction updates. The key you use (e.g., myTxSubscription) is a user-defined label for this specific filter configuration.If all fields are left empty, all transactions are broadcasted. Otherwise:
  • Fields operate as a logical AND.
  • Values within arrays are treated as a logical OR (except for accountRequired, where all must match).
Define filters for block updates. The key you use (e.g., myBlockLabel) is a user-defined label for this specific filter configuration.
This functions similarly to Blocks but excludes transactions, accounts, and entries. The key you use (e.g., blockmetadata) is a user-defined label for this subscription. Currently, no filters are available for block metadata—all messages are broadcasted by default.
Subscribe to ledger entries. The key you use (e.g., entrySubscribe) is a user-defined label for this subscription. Currently, there are no filters available for entries; all entries are broadcasted.

Code Examples (LaserStream SDK)

SDK Options

We provide official SDKs for multiple programming languages: For other languages or custom implementations, you can use the Yellowstone gRPC proto files directly to generate gRPC clients for your preferred language.

Troubleshooting / FAQ

A: Performance issues with LaserStream connections are typically caused by:
  • Javascript Client Slowness: The JavaScript client may lag behind when processing too many messages or consuming too much bandwidth. Consider filtering your subscriptions more narrowly to reduce message volume, switch to the LaserStream JavaScript SDK, or try using another language.
  • Limited local bandwidth: Heavy subscriptions can overwhelm clients with limited network bandwidth. Monitor your network usage and consider upgrading your connection or reducing subscription scope.
  • Geographic distance: Long network routes increase latency and packet loss. Use the endpoint closest to your server. For high-latency connections, increase your network read buffer sizes (can improve bandwidth by 5x+):
    To persist across reboots, add to /etc/sysctl.conf:
    Increase the HTTP/2 stream and connection window sizes to 64MB to prevent flow control bottlenecks. Both are required — raising only the stream window leaves the connection-level window as the binding constraint:
  • Client-side processing bottlenecks: Ensure your message processing logic is optimized and doesn’t block the main thread for extended periods.
Debugging Client Lag: To help you debug client, we built a tool to test for the max bandwidth from your node to a Laserstream gRPC server. To use it run:
The output returns the max network capacity between your server and the Laserstream server. At a minimum, you need 10MB/s to subscribe to all transaction data and 80MB/s to subscribe to all account data. We recommend having at least 2x the required capacity for optimal performance.
A: Verify your API key and endpoint are correct and that your network allows outbound gRPC connections to the specified endpoint. Check the Helius status page for any ongoing incidents.
A: Double-check the logical operators (AND/OR) described in the filter sections. Ensure public keys are correct. Review the commitment level specified in your request.
A: Yes, you can define filter configurations under multiple keys (e.g., accounts, transactions) within the same SubscribeRequest object.
A: We don’t implement consumer groups. Instead, LaserStream delivers the same outcomes teams want: resume, replay, and multi-node reliability without a coordination layer (and the latency/overhead that comes with it). We believe consumer groups are not needed for most workloads and they add latency and operational overhead. As an example a single LaserStream gRPC connection can emit up to 10× Solana’s transaction + account data, and most clients subscribe to a small, filtered slice. Using consumer groups in this case burns performance headroom and introduces another point of failure.
A: LaserStream truncates transaction log messages to 10 KB by default for better speed and performance. If you need full logs, connect to a dedicated untruncated endpoint — see Log Truncation for the list.
A: Including a ping field in your initial SubscribeRequest causes LaserStream to silently ignore all subscription filters — only a Pong is returned with zero account, transaction, or slot data. To fix this, remove ping from the initial subscribe request and instead send pings separately via the stream’s sink after the subscription is established. This keeps the connection alive without interfering with your filters.