Global routing and automatic failover
- High-availability clusters across 3 continents and 10+ data centres
- Cross-region redundancy for global reliability
- Instant failover with no DNS propagation delay
Full suite of Solana streaming solutions for speed, completeness, and control
Triton delivers real-time Solana data with predictable performance at any scale. Built on gRPC and engineered to handle high throughput, this is the streaming foundation behind the most demanding systems across the ecosystem.
Choose the right streaming engine for any workload
Riptide and Whirligig for raw speed, Fumarole for persistent gap-free indexing, Deshred for pre-execution transactions, Triton Shred Streaming for a blended multi-source shred feed, Account Sync for streaming benefits with polling code.
Riptide gRPC streams, Dragon's Mouth compatible
Riptide is how Triton serves the Dragon's Mouth interface, over a faster transport. Same protocol, same subscribe request, same filters, same client code: you point your client at a new endpoint URL and nothing else changes.
- Same wire protocol as Dragon's Mouth, so no client code changes
- Intra-slot data through a Geyser-fed gRPC interface, with minimal serialisation overhead
- Scales to millions of accounts per connection with compressed filters
- Built-in auto-reconnect with deduplicated replay over a ~1,000-slot buffer
Best for traders, market makers, arbitrage, and bots. Pair with Deshred for pre-execution intent.
Coming from Dragon’s Mouth? Riptide speaks the same protocol, so your client works unchanged. Swap your endpoint URL and you are on it. Riptide runs on every Triton endpoint, shared and dedicated, at the same price.
Riptide gRPC streams, Dragon's Mouth compatible
Riptide is how Triton serves the Dragon's Mouth interface, over a faster transport. Same protocol, same subscribe request, same filters, same client code: you point your client at a new endpoint URL and nothing else changes.
- Same wire protocol as Dragon's Mouth, so no client code changes
- Intra-slot data through a Geyser-fed gRPC interface, with minimal serialisation overhead
- Scales to millions of accounts per connection with compressed filters
- Built-in auto-reconnect with deduplicated replay over a ~1,000-slot buffer
Best for traders, market makers, arbitrage, and bots. Pair with Deshred for pre-execution intent.
Coming from Dragon’s Mouth? Riptide speaks the same protocol, so your client works unchanged. Swap your endpoint URL and you are on it. Riptide runs on every Triton endpoint, shared and dedicated, at the same price.
Whirligig WebSocket streams
Browsers don't support gRPC directly, but UI workloads still require live chain updates. Whirligig bridges this gap by wrapping Dragon's Mouth engine in an enhanced WebSocket interface optimised for the web, allowing you to track balances updates and confirm transactions milliseconds after they happen on-chain.
- Drop-in for Solana's native WebSocket API
- Full blockSubscribe and transactionSubscribe method support
- Powered by Dragon's Mouth gRPC
- All WebSocket connections routed through it by default
Best for browsers, dApps, dashboards, and real-time feeds.
Whirligig WebSocket streams
Browsers don't support gRPC directly, but UI workloads still require live chain updates. Whirligig bridges this gap by wrapping Dragon's Mouth engine in an enhanced WebSocket interface optimised for the web, allowing you to track balances updates and confirm transactions milliseconds after they happen on-chain.
- Drop-in for Solana's native WebSocket API
- Full blockSubscribe and transactionSubscribe method support
- Powered by Dragon's Mouth gRPC
- All WebSocket connections routed through it by default
Best for browsers, dApps, dashboards, and real-time feeds.
Fumarole persistent streams
Fumarole is a high-availability persistent streaming system that handles gap detection and backfill automatically on every reconnect. It tracks your position in the stream, so you can resume exactly where you left off, days after a drop. Multiple gRPC nodes feed into a single consistent timeline, so you only receive complete and correct data.
- Maximum redundancy through multiplex Geyser ingestion
- 4-day rolling cache for historical replay
- Server-side cursor memory of what data was delivered
- At-least-once delivery guarantee
- Three streaming modes: slot-sequential, block stream, and raw
Best for indexers, analytics, and workloads that rely on data completeness.
Fumarole persistent streams
Fumarole is a high-availability persistent streaming system that handles gap detection and backfill automatically on every reconnect. It tracks your position in the stream, so you can resume exactly where you left off, days after a drop. Multiple gRPC nodes feed into a single consistent timeline, so you only receive complete and correct data.
- Maximum redundancy through multiplex Geyser ingestion
- 4-day rolling cache for historical replay
- Server-side cursor memory of what data was delivered
- At-least-once delivery guarantee
- Three streaming modes: slot-sequential, block stream, and raw
Best for indexers, analytics, and workloads that rely on data completeness.
Triton Shred Streaming
Triton Shred Streaming delivers shreds pulled directly from Triton's own validators and blended with third-party sources into a single feed. You receive whichever copy of each shred arrives first, from whichever source saw it first, and reconstruct transactions on your own hardware.
- Shreds direct from Triton's own high-stake validators, blended with DoubleZero and other sources
- Whichever source delivers a given shred first is the one you get
- A shred delayed or dropped by one source still reaches you through another
Best for HFT desks, arbitrage bots, and anyone who thinks 1 millisecond is too long to wait.
Triton Shred Streaming
Triton Shred Streaming delivers shreds pulled directly from Triton's own validators and blended with third-party sources into a single feed. You receive whichever copy of each shred arrives first, from whichever source saw it first, and reconstruct transactions on your own hardware.
- Shreds direct from Triton's own high-stake validators, blended with DoubleZero and other sources
- Whichever source delivers a given shred first is the one you get
- A shred delayed or dropped by one source still reaches you through another
Best for HFT desks, arbitrage bots, and anyone who thinks 1 millisecond is too long to wait.
Deshred transactions
Deshred delivers the earliest possible signal of on-chain activity. It reconstructs transactions from raw shreds and streams them before the node processes them, skipping the replay stage and giving traders the lowest-latency view of intent on Solana.
- p50 ~6.3 ms, p90 ~20 ms faster than standard gRPC
- Server-side vote and account filtering
- Full transaction reconstruction with ALT resolution
- Same gRPC subscription pattern as Dragon's Mouth
Best for arbitrage, MEV, market making, copy trading, and HFT.
Deshred transactions
Deshred delivers the earliest possible signal of on-chain activity. It reconstructs transactions from raw shreds and streams them before the node processes them, skipping the replay stage and giving traders the lowest-latency view of intent on Solana.
- p50 ~6.3 ms, p90 ~20 ms faster than standard gRPC
- Server-side vote and account filtering
- Full transaction reconstruction with ALT resolution
- Same gRPC subscription pattern as Dragon's Mouth
Best for arbitrage, MEV, market making, copy trading, and HFT.
Account Sync
A drop-in SDK that resolves your RPC reads from a local in-memory cache, kept fresh by a streaming subscription on Triton's side. You get streaming-grade speed and cost on getAccountInfo and getMultipleAccounts without rewriting your polling code.
- One-line SDK swap (e.g. for @solana/kit)
- Account reads resolved from local RAM, instead of the network
- Bandwidth-only billing at $0.08/GB, no per-request charges
- Bounded queues with newest-first overflow during traffic bursts
Best for wallets, DEXs, portfolio trackers, NFT marketplaces, and any app polling RPC at scale.
Account Sync
A drop-in SDK that resolves your RPC reads from a local in-memory cache, kept fresh by a streaming subscription on Triton's side. You get streaming-grade speed and cost on getAccountInfo and getMultipleAccounts without rewriting your polling code.
- One-line SDK swap (e.g. for @solana/kit)
- Account reads resolved from local RAM, instead of the network
- Bandwidth-only billing at $0.08/GB, no per-request charges
- Bounded queues with newest-first overflow during traffic bursts
Best for wallets, DEXs, portfolio trackers, NFT marketplaces, and any app polling RPC at scale.
Co-located shred ingestion
- Co-located with high-stake validators
- Jito, Double Zero, and Triton shreds raced for fastest delivery
- Earliest availability stream, delivered by default
Level playing field
- No backrunning, frontrunning, or sandwiching
- No private orderflow or other value extraction from your traffic
- Equal access for every builder, with free SWQoS on request
Built on open standards
- Open-source stack across the streaming suite
- No vendor lock-in or closed-source APIs
- Industry-standard architecture used across the entire ecosystem
Direct support channel on every endpoint
- Access to the top-tier engineering team
- Configuration help, debugging, and urgent fixes
- Reachable 24/7 across all the regions
Intelligent abuse prevention
- Access controls and origin allowlists
- JSON-RPC validation and request fingerprinting
- No tier-gated throttling that blocks legitimate workloads