Skip to content
ZecFlow

ZecFlow

Build Private Agentic Web3 Workflows with Nillion

Created on 4th December 2025

ZecFlow

ZecFlow

Build Private Agentic Web3 Workflows with Nillion

The problem ZecFlow solves

ZecFlow gives teams the exact blocks they need to build private workflows; NilAI for private inference, NilCC for confidential computation, NilDB for private storage, Zcash blocks for private value transfer—inside a drag-and-drop builder so they can launch private workflows without wiring every primitive by hand.

image

That visual pipeline solves three chronic problems:

1. Privacy vs. productivity: normally you either ship fast in plaintext or stall while inventing bespoke encryption; ZecFlow keeps state encrypted by default so teams stay fast and compliant through our visual builder.

2. Opaque AI/compute: regulators and users want proof that secure hardware actually ran the job; NilAI + NilCC inject attestations and signatures that the UI can surface on every run.

3. Painful selective disclosure: sharing results usually means passing around keys or dumping data; NilDB delegation lets operators expose only the final decision (diagnosis, loan verdict, etc.) through short-lived reads.

Because those guarantees are baked into the same workflow editor, teams can prototype medical, lending, or other sensitive automations while confidently stating that inputs, intermediate values, and outputs remain shielded unless policy allows a reveal.

Challenges I ran into

1. NilAI attestation failures (401s): the inference client initially fetched attestations with the raw API key, so NilAI returned

undefined

fields that blew up NilDB encryption. Reusing the NilAuth invocation token and caching the attestation response stabilized the flow.

2. NilDB

DataValidationError:

must be object: NilDB standard data only accepts encrypted primitives, but NilAI responses shipped nested

raw

/

attestation

blobs. I wrote a sanitizer that flattens the payload into strings/hashes before encryption and reconstructed the attestation from NilAI directly when displaying results.

3. No metal instance for NilCC workload: the loan workflow requires NilCC metal capacity, but the public testnet had no free slots, so I added guardrails (UI warning, graceful error capture, and builder status logs) that keep the rest of the demo functional while clearly telling users the NilCC path may fail if no metal instance is available.

Tracks Applied (6)

Privacy-Preserving AI & Computation

ZecFlow gives teams the exact blocks they need to build private workflows; NilAI for private inference, NilCC for confid...Read More
Axelar Network

Axelar Network

Privacy Infrastructure & Developer Tools

ZecFlow gives teams the exact blocks they need to build private workflows; NilAI for private inference, NilCC for confid...Read More

Electric Coin Company

Privacy Infrastructure & Developer Tools

ZecFlow gives teams the exact blocks they need to build private workflows; NilAI for private inference, NilCC for confid...Read More

Zcash Community Grants

Privacy-Preserving AI & Computation

ZecFlow enables end-to-end privacy-preserving workflows on the Nillion stack: NilAI provides signed, attestable inferenc...Read More
nillion

nillion

Privacy Infrastructure & Developer Tools

ZecFlow gives teams the exact blocks they need to build private workflows; NilAI for inference, NilCC for computation, N...Read More

Raybot

General Bounty

ZecFlow gives teams the exact blocks they need to build private workflows; NilAI for inference, NilCC for computation, N...Read More

Project Tachyon

Discussion

Builders also viewed

See more projects on Devfolio