For AI agents: the documentation index is at /llms.txt. Markdown versions of pages are available by appending .md to the URL.
Skip to main content

Indexing on Solana

HyperIndex indexes Solana programs at the instruction level. You select the programs and instructions you care about, HyperIndex decodes them (arguments and accounts) using your Anchor IDL or an inline schema, and writes the results to Postgres with an auto-generated GraphQL API. Inner instructions (CPIs), token balances and balance changes, transaction metadata and program logs are all available.

It is powered by HyperSync for Solana, the same high-performance data engine behind EVM indexing, so historical backfills are fast and you never touch an RPC node for the bulk of indexing.

Beta

Solana support is in beta: the core instruction-indexing path is stable enough to build on, though the program/instruction config surface (the chain-level experimental key) may still evolve as we round it out. Building on Solana? Say hello on Discord. We'd love your input on what to prioritize next.

Two ways to index Solana​

ApproachAPIData sourceUse it for
Instruction handlersindexer.onInstructionHyperSyncThe main path: decode and index program instructions (swaps, deposits, mints, transfers…), including inner/CPI instructions, with token balance changes.
Slot handlersindexer.onSlotRPC (via the Effect API)Per-slot orchestration, time-series snapshots, or pulling extra data from RPC on a schedule.

Most indexers use instruction handlers. Slot handlers are for cases where you need to run logic on a slot cadence rather than react to a specific instruction. For raw, low-level data you can also query HyperSync for Solana directly.

Quickstart​

pnpx envio init

Choose Solana when prompted, then pick a starter template (a Metaplex NFT instruction indexer, or a minimal slot handler). See Getting Started for the full walkthrough.

Data endpoint and history​

Every Solana chain must name its HyperSync endpoint explicitly under experimental.hypersync_config.url - it is a required field, with no default applied when it is missing. Each endpoint serves history back to its own floor slot, and that floor moves forward over time, so pick both the endpoint and start_block deliberately. See choosing an endpoint for the current endpoints and the silent-skip failure mode to avoid.

Mental model: coming from EVM?​

If you've used HyperIndex on EVM, the shift is mostly vocabulary:

EVMSolana
Contract + ABIProgram + IDL
Event (onEvent)Instruction (onInstruction)
event.paramsinstruction.params?.args (optional: decoding can fail)
Topic0 / event signatureInstruction discriminator
Block (onBlock)Slot (onSlot)
Hex addresses 0x…Base58 addresses
start_block = block numberstart_block = slot number

See EVM vs Solana for the full picture.

What's supported today​

  • Instruction indexing via indexer.onInstruction: match by program + discriminator.
  • IDL-aware decoding: point at a standard Anchor IDL (legacy or 0.30+) and HyperIndex derives the argument and account layout. The discriminator is always read from config.yaml, never from the IDL. No IDL? Declare an inline schema.
  • Inner instructions (CPIs): decoded the same way as top-level ones, with a full instruction-address path so you can reconstruct the call tree.
  • Token balances & balance changes: pre/post SPL Token (and Token-2022) balances per transaction, so you get the net token movement without indexing every transfer. See token balances.
  • Transaction metadata & logs: fee payer, fee, compute units, success, the transaction signature, and per-instruction program logs (opt-in via field selection).
  • Slot handlers via indexer.onSlot + the Effect API for RPC enrichment.
  • Local dev + GraphQL + Envio Cloud: the same workflow and hosting as EVM.

What is not supported yet​

These are gaps in the built-in instruction-handler surface, not hard limits: for most of them you can still pull the data yourself by calling out to RPC from a handler with the Effect API, you just don't get it as a struct field for free.

  • Native SOL balance fields on handlers. Pre/post token balances are surfaced today; native SOL (lamport) balances aren't yet a handler field-selection toggle, though you can get them via HyperSync directly, or by reading the account over RPC in a handler with the Effect API.
  • Account-change subscriptions. There is no onAccount/program-account handler. instruction.accounts and params.accounts give you the account references (pubkeys) an instruction touched, not their state; for actual state, per-transaction token balances (via token_balance_fields) cover the token case, and the Effect API covers everything else by reading it over RPC yourself.
  • A separate log handler. Logs are a field on the instruction event, not their own handler.
  • No-code contract import. Solana has no contract-import flow, so you configure programs/instructions by hand. (IDLs are wired up in config.yaml, not auto-imported.)
  • ReScript. Solana indexers are TypeScript only. Codegen emits no ReScript for ecosystem: svm, and envio init silently picks TypeScript if you ask for ReScript.
  • Per-field selection on token_balance_fields / log_fields. Those two toggles accept true only. transaction_fields and block_fields do take field name lists.

If the piece you need is on this list, tell us on Discord: there's a good chance we can sequence the work to unblock you, or point you at a HyperSync-direct path that gets the data today.

In this section​