If you've heard terms like "Geyser plugin," "Yellowstone gRPC," or "Dragon's Mouth" and wondered what they mean — this is the guide for you. Geyser is the foundation of real-time Solana data, and understanding it explains why gRPC-based bots have a structural speed advantage.

Geyser in 30 Seconds

Geyser is a plugin interface built into the Solana validator (agave-validator). When a validator processes a slot, Geyser hooks fire synchronously with the validator's execution. Data goes directly from validator memory to the plugin — zero network hops, zero serialization delay.

The Four Callback Types

Callback	Fires When	Data Provided
`on_account_update`	Any account changes state	Account pubkey, owner, data, lamports
`on_transaction`	A transaction is confirmed	Full transaction with all instructions
`on_slot_status`	Slot status changes	Slot number, status (processed/confirmed/finalized)
`on_block_metadata`	Block is completed	Block hash, rewards, timestamps

Why This Is Faster Than RPC

Traditional RPC works like this: your bot asks "what happened?" and the RPC node queries its stored data and responds. This introduces querying latency, serialization overhead, and polling gaps.

Geyser works like this: data is pushed to the plugin the instant it exists in validator memory. No query, no storage round-trip, no polling interval. The data flows: Validator Memory → Geyser Hook → Plugin → Your Bot.

Available Geyser Plugins

Yellowstone gRPC (Dragon's Mouth)

The most popular plugin, maintained by Triton One. Streams data to connected clients via gRPC (HTTP/2 binary protocol). Used by QuickNode, Chainstack, Solana Tracker, and Subglow. Client libraries available for TypeScript, Rust, Python, and Go.

Yellowstone Kafka

Streams Geyser data to Apache Kafka topics. Designed for high-throughput data pipelines where you need replay capabilities and multiple consumers.

Yellowstone PostgreSQL

Writes Geyser data to a PostgreSQL database. Used for historical queries, analytics, and indexing. Not real-time streaming — optimized for storage and querying.

Yellowstone Vixen

A framework for building custom parsers on top of Geyser data. Useful when you need program-specific parsing logic without building a full plugin.

Geyser vs RPC vs WebSocket

Feature	Geyser (gRPC)	JSON-RPC	WebSocket
Data path	Validator memory → plugin → client	Validator → storage → HTTP response	Validator → pubsub → notification
Latency	~5ms	200-800ms	50-200ms
Data completeness	Everything (full transactions)	What you query	Notifications only
Filtering	Server-side, before transmission	None	Limited (account subs)
Missed data	None — persistent stream	Common during congestion	Possible on disconnect

Running Geyser vs Using a Provider

Running your own Geyser setup: Requires a Solana validator or RPC node ($500-2,000/mo hardware), Yellowstone plugin configuration, DevOps expertise, and 24/7 monitoring. Best for institutional traders with extreme latency requirements.

Using a provider: Managed endpoint, global regions, uptime guarantees, and in Subglow's case, pre-parsed output. Best for most teams — faster to start, lower cost, professionally managed.

How Subglow Uses Geyser

Subglow runs Yellowstone gRPC on professionally managed validators and adds two layers on top:

Server-side program filtering — only Pump.fun, Raydium, and Jupiter events reach your bot
Pre-parsed JSON output — Borsh decoding happens on our infrastructure, you get clean JSON

Use the standard @triton-one/yellowstone-grpc client or any gRPC client. Point it at grpc.subglow.io and you're streaming.

Related Guides

Yellowstone gRPC Tutorial — hands-on connection guide
Yellowstone gRPC Filters — complete filter reference
gRPC Provider Comparison — compare providers and pricing
Solana gRPC vs RPC — detailed protocol comparison

What Is Solana Geyser? The Plugin Framework That Powers Real-Time Data