No single open source Solana shred parser library covers the whole path from packet to program event. The official Agave crates handle shreds and entries, the solana-sdk crates handle transactions, and Python, Go and TypeScript libraries parse transactions but not shreds. Expect to combine two or three libraries and write the program decoding yourself.
Layers a parser can cover#
- Shred header: reading the slot, index, type and erasure set fields from one datagram.
- Recovery: rebuilding missing data pieces from the erasure set.
- Entries: turning the recovered byte stream into entries and block markers.
- Transactions: telling legacy, v0 and v1 apart and parsing each one.
- Lookup tables: reading table account data so v0 addresses can be resolved.
- Program instructions: decoding each program's instruction data, from an IDL or from source.
The cards below say which of these each library handles. None of them does layer 6 for you in general.
The libraries#
solana-ledger (Rust)#
Repository: anza-xyz/agave. Licence: Apache-2.0. Stable version used by our docs: 4.4.1. Covers layers 1 to 3: shred types and the shredder. It only exposes these with the agave-unstable-api feature, and building it from crates.io needs the same patched dependencies Agave itself pins. Its built-in recovery path expects a validator bank, so standalone receivers usually recover with a separate erasure crate (E4 tech facts C-01).
solana-entry (Rust)#
Repository: anza-xyz/agave. Licence: Apache-2.0. Version 4.4.1. Covers layer 3: the Entry type and block components. Also behind agave-unstable-api, but it does not pull in the database dependencies (E4 tech facts C-02).
reed-solomon-erasure (Rust)#
Repository: darrenldl/reed-solomon-erasure. Licence: MIT. Version 6.0.0. Covers layer 2 when paired with your own shred grouping. It knows nothing about Solana; you feed it the shards of one erasure set.
solana-transaction and solana-message (Rust)#
Repository: anza-xyz/solana-sdk. Licence: Apache-2.0. Latest version 6.0.0, but pair solana-entry 4.4.1 with the 5.x line it depends on, or the transaction types will not match (E4 tech facts C-01). Covers layer 4 for all three live formats, legacy, v0 and v1 (E4 tech facts C-03, C-04, T-01).
solana-address-lookup-table-interface (Rust)#
Repository: anza-xyz/solana-sdk. Licence: Apache-2.0. Version 4.0.0. Covers layer 5: the account layout of a lookup table, so you can read the addresses once you have fetched the account (E4 tech facts C-05, T-07).
solana-transaction-status (Rust)#
Repository: anza-xyz/agave. Licence: Apache-2.0. Version 4.4.1. Defines the execution metadata RPC returns. Useful when you reconcile your decoded rows with RPC; it has nothing to parse in shreds, because shreds carry no execution data (E4 tech facts C-06, X-02).
solders (Python)#
Repository: kevinheavey/solders. Licence: Apache-2.0. Version 0.29.0. Covers layer 4: VersionedTransaction.from_bytes parses legacy, v0 and v1 messages (E4 tech facts C-07). Shred handling is not part of it.
solana-go (Go)#
Repository: now under the solana-foundation organisation on GitHub; the Go module path at v1.26.0 is still github.com/gagliardetto/solana-go. Licence: Apache-2.0. Version v1.26.0. Covers layer 4 for legacy, v0 and v1. One trap: its message version getter returns an internal enum, where v1 is the value 2, so compare against the library's constant instead of printing the number (E4 tech facts C-08).
@solana/kit (TypeScript)#
Package: @solana/kit on npm. Licence: MIT. Version 8.6.0 at the check date. Covers transaction decoding in JavaScript runtimes. Tested at 8.6.0 against our full fixture set, it handled all three formats correctly (E4 tech facts C-09).
Coverage grid#
| Library | 1 Header | 2 Recovery | 3 Entries | 4 Transactions | 5 Lookup tables | 6 Programs |
|---|---|---|---|---|---|---|
| solana-ledger | Yes | Needs a bank | Yes | No | No | No |
| solana-entry | No | No | Yes | Uses solana-transaction | No | No |
| reed-solomon-erasure | No | Yes, generic | No | No | No | No |
| solana-transaction, solana-message | No | No | No | Yes, all three formats | No | No |
| solana-address-lookup-table-interface | No | No | No | No | Yes | No |
| solders | No | No | No | Yes, all three formats | No | No |
| solana-go | No | No | No | Yes, all three formats | No | No |
| @solana/kit | No | No | No | Yes, all three formats at 8.6.0 | No | No |
Reuse or write your own#
For layers 1 to 3 the Rust crates are the reference: they are the code Agave validators run. If your stack is not Rust, you have two realistic choices. Either run a small Rust stage that turns shreds into transaction bytes and hands them to the rest of your system, or port the shred and entry parsing yourself and test it against the Rust output.
For layer 4, reuse is the clear answer in every language listed. Writing your own transaction parser is only worth it when you need something the libraries do not offer, such as zero allocation parsing on a hot path. If you do, fuzz it: your decoder will meet garbage explains why.
Layer 6 has no general library. You decode each program from its IDL or its source, which is what the Rust decoder guide and the program recipes build towards.
Keeping up with network upgrades#
Formats move. Transaction v1 became active on mainnet at epoch 1035 (E4 tech facts T-05), and a parser written before that date will reject or misread v1 transactions. Two habits help:
- Pin library versions, and upgrade them deliberately with fixture tests that include all three transaction formats.
- Watch the Solana Improvement Documents for changes to transaction and shred formats, and the Agave release notes for the crates.
The versioned transactions explainer shows the first byte test that tells the formats apart, which is the place most older parsers break.
Questions about parser libraries#
Is there a single crate that turns shreds into transactions?#
Not as a supported public API. The Agave crates contain the pieces, but you assemble them yourself and compile with the unstable feature flag.
Can I parse shreds in Python or Go?#
The transaction libraries in those languages do not parse shreds. Most teams either port the shred and entry parsing or run a Rust stage in front.
Why does the version matter so much?#
Because v1 transactions are live. A library that predates v1 support will fail on them, and silently dropping a transaction format is worse than a crash.
Are seller-supplied decoders included here?#
No. Some sellers bundle their own decoders with their feeds; this page lists only open source libraries you can read and pin yourself.
What this catalogue does not tell you#
It records versions and licences as published on the check date, not quality or performance. It does not cover community crates without a stable release. Licences can change between versions, so read the licence file of the exact version you pin.
Next steps#
Pick a language with the stage by stage language comparison, see how the libraries fit together in the decoding pipeline, and check what each stream message carries in the message format comparison. When you are ready to point a library at live packets, start with the quickstart.
Sources#
- crates.io: solana-ledger (Apache-2.0, repository anza-xyz/agave) (checked 2026-11-18)
- crates.io: solana-entry (Apache-2.0) (checked 2026-11-18)
- crates.io: solana-transaction 6.0.0 (Apache-2.0, repository anza-xyz/solana-sdk) (checked 2026-11-18)
- crates.io: solana-message 6.0.0 (Apache-2.0) (checked 2026-11-18)
- crates.io: solana-address-lookup-table-interface 4.0.0 (Apache-2.0) (checked 2026-11-18)
- crates.io: solana-transaction-status (Apache-2.0) (checked 2026-11-18)
- crates.io: reed-solomon-erasure 6.0.0 (MIT) (checked 2026-11-18)
- solders on GitHub (Apache-2.0); 0.29.0 on PyPI (checked 2026-11-18)
- solana-go on GitHub (Apache-2.0), v1.26.0 (checked 2026-11-18)
- @solana/kit on npm (MIT) (checked 2026-11-18)
Raw shreds on your own server
Ultra Low Latency Shreds is $200 and Standard Shreds is $100. You pay in crypto, tell us which IPv4 address and UDP port should receive the stream, and we wire it up by hand.