Turbo Kaldera (aRPC) — Pre-Execution Solana Streaming

Turbo Kaldera is constant-k's aRPC — anticipatory RPC. It speaks the same Yellowstone protocol as Kaldera, but the aRPC stream is built from transaction data captured and reassembled before the validator executes it. You see a transaction the moment it physically exists on the wire — not after the chain has finished processing it.

What is aRPC?

Anticipatory, Not Reactive

A Solana transaction passes through five stages: submitted, packed into shreds, reassembled, executed, committed. Every standard RPC and gRPC stream fires after execution. aRPC — anticipatory RPC — serves the stream at stage three: the instant shreds are reassembled back into transactions, before the bank has run them. Two stages earlier than the data everyone else is trading on.

Managed Reconstruction

Raw shred feeds hand you UDP packets and make you do the work — reassembly, dedup, decode, all on your latency budget. Turbo Kaldera does the reconstruction server-side and hands you the aRPC stream over the standard Yellowstone protocol. The earliest data, none of the plumbing.

Drop-In Yellowstone

Same protocol, same Subscribe call, same client library: point your existing Yellowstone client at the turbo endpoint (turbo-kaldera-newyork / turbo-kaldera-frankfurt) instead of the standard one, and keep your code. Account include/exclude filters work exactly as before — account keys exist in the transaction message pre-execution.

Raw and Unconfirmed by Design

This is the tip of the data, and we want you to use it knowing exactly what it is. Pre-execution transactions carry a real signature, message and account keys — but no fee, logs, balance changes or success/failure, because those do not exist until execution. Built for signature- and account-driven strategies; pair it with standard Kaldera when you need outcomes.

Turbo or Standard?

Run turbo when being first matters more than being complete: copy trading, MEV detection, watching a pool for the first touch. Run standard Kaldera when your logic reads execution results. Many subscribers run both — turbo to react, standard to confirm.
Turbo Kaldera runs in both New York and Frankfurt alongside standard Kaldera, using the same plans and API keys — see the pricing page and plan limits. Start with standard Kaldera when you need execution results, add the turbo endpoint when you need to be first. Raw shreds instead? That is ShredPrism.