How Privy Powers Onchain Context for 120+ Million Wallets
Henri Stern, CEO & Co-Founder, on building wallet infrastructure, real-time streaming, and the data layer for agentic payments
Privy builds the agentic wallet layer for the internet. Developers embed self-custody wallets directly into their products through Privy's API — no browser extension or seed phrase required. More than 2,000 developer teams have built on top of it. The result is 120+ million accounts powering some of the most-used applications in crypto and fintech: Ramp, Deel, Hyperliquid, and OpenSea, among others.
In June 2025, Stripe acquired Privy to anchor the wallet layer of its crypto stack and Bridge to build out stablecoin infrastructure. Alongside Paradigm, Stripe also incubated Tempo, a Layer-1 blockchain purpose-built for stablecoin and real-world payments. Together they form a programmable money layer: wallets that hold and manage assets, stablecoin rails that move them, and payments infrastructure to orchestrate it all.
Henri Stern is Privy's co-founder and CEO. He's been building wallet infrastructure since before most companies took the problem seriously. We asked him how he thinks about the data layer that sits underneath every wallet Privy powers.
A wallet without context is just a key
The common instinct is to think of wallets as key management systems: generate a private key, secure it, sign transactions.
But that's not how Henri thinks about it:
"Wallets are not just key management systems. They're also policy engines and [venues for] onchain execution, and all of this has to be aware of onchain context. You're not making decisions in a vacuum. You're making decisions based on what's happening with the systems you're reacting to."
That context requirement runs through every layer of Privy's product. When a user checks their balance, Privy needs the current onchain state. When a transfer completes, the developer building on Privy needs to know, so they can update the in-app experience their user sees in real time. When an agent holds a wallet and decides whether to make a payment, it needs accurate, current data to make that decision correctly.
Privy's data architecture, built on Allium
Privy has two main types of data requirements:
Historical and analytical workloads (balance APIs, internal reconciliation, in-app transaction history) run through Allium's Snowflake secure datashare. The underlying data covers decoded transactions across 14+ EVM chains plus Solana, TRON, and Hyperliquid, landing directly in Privy's data environment.
Real-time, customer-facing workloads also run through Allium. When an end user makes a transfer inside a Privy-powered app, the sequence looks like this:
- An onchain event occurs
- Allium streams it into Privy's Kafka instance in real time
- Privy fires a webhook to the developer
- The developer's app updates the user's screen
Privy needed a streaming layer where webhook gaps that fail without surfacing errors are structurally off the table, not just unlikely. That meant custom Kafka configuration, backfill with up to 7-day replay windows, and testnet parity across every mainnet chain for staging. After more than a month of parallel testing with no incidents, Privy moved production traffic to Allium.
What Allium enables for Privy
Wallet state users can trust. Users of Privy-powered apps see accurate, current wallet state — balances, transaction history, whether something went through. Allium reads and indexes onchain data across chains so Privy can surface that reliably without building chain-by-chain indexing infrastructure themselves.
Apps that react when transactions complete. When a transaction completes onchain, the developer building on Privy needs to know so their app can react. Allium pipes those events in real time.
AI agents can act on current onchain state. Agents acting autonomously need to know what's actually happening onchain to make informed decisions. At the scale of thousands of agents running in parallel, a data quality issue doesn't affect one transaction — it propagates across every decision the fleet makes.
Privy's emerging use case: data for agents
Privy is already being used to arm AI agents with wallets, providing them the means to hold funds, operate within policy guardrails, and transact on behalf of users without human intervention. But a wallet, by itself, isn't enough for an agent to act well.
"What's missing in all of this is the data layer. How is an agent to make informed decisions? It's one thing when there's a human in the loop who can always interrupt the workflow, correct things, see that something is off. But if you're trying to have these non-deterministic systems make rapid autonomous decisions — and you're running a fleet of thousands of them — any issue with the data quality you're getting just ripples and is so far worse."
Henri frames the agentic stack as three foundational layers that together enable autonomous systems to act on the web in a contextual way:
- Privy: the wallet and control plane. Policy engines that define what the agent can spend, on what, and under what conditions. If something goes wrong, guardrails limit the damage.
- Allium: the data layer. Onchain context so agents make decisions grounded in what's actually happening, not stale or incomplete state.
- The payment mechanism: protocols like x402 and MPP let agents access data and services autonomously, without pre-negotiated contracts or human intermediaries.
"The marriage of Privy, Allium, and a payment protocol like x402 or MPP is the legs of the stool for enabling autonomous systems that can act on the web in a contextual way."
Learn more about Privy. Follow Henri Stern on X at @sternhenri.
Building wallet infrastructure, real-time data pipelines, or agent-native financial applications? Explore Allium ->
