EZKL
ZkonduitEZKL turns a neural network exported to ONNX into a zero-knowledge circuit and proves that an inference was run on that model, without revealing the inputs or weights the prover chooses to keep private. Proofs use the Halo2 proof system and verify in a browser, on a device or on-chain in the Ethereum Virtual Machine. Zkonduit maintains it as a Rust library with Python bindings, and publishes a JavaScript engine package on npm.
Tagged independently verifiable: a verifier checks the proof itself. The repository carries no license grant, so the code can be read but not freely reused.
Openness
2 high confidence- license
- none-declared(no LICENSE file, no license field in Cargo.toml or pyproject.toml
- source
- public(Rust library and Python bindings)
The full source is public, but the repository grants no license: there is no LICENSE file and neither package manifest names one, so by default all rights stay with the authors. That makes EZKL readable rather than open source, whatever its contributor agreement says about incoming code. No license is declared anywhere the publisher could declare one, so the license tier is unstated.
- https://raw.githubusercontent.com/zkonduit/ezkl/main/Cargo.toml recorded 2026-09-26
Package manifest for ezkl with no license or license-file field
- https://raw.githubusercontent.com/zkonduit/ezkl/main/cla.md recorded 2026-09-26
Contributor License Agreement covering what contributors grant for their contributions; it licenses no rights in Zkonduit's own code
- https://raw.githubusercontent.com/zkonduit/ezkl/main/pyproject.toml recorded 2026-09-26
Python build manifest with no license field
- https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zkonduit%2Fezkl recorded 2026-09-26
zkonduit/ezkl: license null, archived false, fork false, pushed 2026-02-20
- https://ungh.cc/repos/zkonduit/ezkl/files/main recorded 2026-09-26
Full file listing of zkonduit/ezkl main, 545 paths; no LICENSE, COPYING or LICENSES/ path
Adoption
1 medium confidenceDownloads of the product's own packages measure its use: the Python package, its GPU build and the JavaScript engine on npm.
- https://api.npmjs.org/downloads/point/last-month/@ezkljs/engine recorded 2026-09-26
979 downloads for @ezkljs/engine, 2026-08-27 to 2026-09-25
- https://pypistats.org/api/packages/ezkl-gpu/recent recorded 2026-09-26
last_month downloads = 13 for ezkl-gpu
- https://pypistats.org/api/packages/ezkl/recent recorded 2026-09-26
last_month downloads = 1,824 for ezkl
Capability
5 medium confidenceA proof lets anyone confirm that a specific model produced a specific output without trusting the party that ran it, which puts EZKL beside NVIDIA's hardware attestation at the top of the category.
- https://docs.ezkl.xyz/ recorded 2026-09-26
EZKL documentation site for the library and its bindings
- https://raw.githubusercontent.com/zkonduit/ezkl/main/README.md recorded 2026-09-26
"we use the collaboratively-developed Halo2 as a proof system. The generated proofs can then be verified with much less computational resources, including on-chain (with the Ethereum Virtual Machine), in a browser, or on a device"
Verified 2026-09-26